CN113538155A

CN113538155A - Data processing method and device based on system dynamics model and computer equipment

Info

Publication number: CN113538155A
Application number: CN202011500469.3A
Authority: CN
Inventors: 郑建光; 王硕佳; 顾钰璇; 刘亚飞
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-12-17
Filing date: 2020-12-17
Publication date: 2021-10-22

Abstract

The application relates to a data processing method, a data processing device, computer equipment and a storage medium based on a system dynamics model, and an input data set corresponding to a service system to be analyzed is obtained; the input data set comprises prediction related data and adjustment related data; inputting the input data set into a system dynamic model corresponding to the service system to be analyzed; the system dynamics model comprises a plurality of subsystem models with system dynamic connection, and each subsystem model is a causal relationship network established based on the causal relationship of the prediction correlation data and the adjustment correlation data; respectively carrying out data processing on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result; the service simulation result comprises a data processing result obtained by processing data by different subsystem models. By adopting the method, the accuracy and the efficiency of data processing can be improved.

Description

Data processing method and device based on system dynamics model and computer equipment

Technical Field

The present application relates to the field of computer technologies, and in particular, to a data processing method and apparatus based on a system dynamics model, a computer device, and a storage medium.

Background

With the development of computer technology, various complex big data systems, such as equipment management systems, medical insurance systems, endowment insurance systems, and the like, have appeared.

In the conventional technology, for a big data system involving a large amount of associated data, an acquisition part generally performs data processing on the associated data having a direct influence on a result to obtain a data processing result. However, only one-sided data processing results can be obtained by performing data processing based on the directly affected part of the associated data, and the accuracy of the data processing results is low.

Disclosure of Invention

In view of the above, it is necessary to provide a data processing method, an apparatus, a computer device and a storage medium based on a system dynamics model, which can improve the accuracy of data processing.

A method of data processing based on a system dynamics model, the method comprising:

acquiring an input data set corresponding to a service system to be analyzed; the input data set comprises prediction related data and adjustment related data;

inputting the input data set into a system dynamic model corresponding to the service system to be analyzed; the system dynamics model comprises a plurality of subsystem models with system dynamics connections; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of prediction associated data and adjustment associated data;

respectively carrying out data processing on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result; the service simulation result comprises a data processing result obtained by processing data by different subsystem models.

In one embodiment, obtaining an input data set corresponding to a service system to be analyzed includes:

displaying a resource analysis interface corresponding to the service system to be analyzed; displaying candidate associated data on the resource analysis interface; and receiving data adjusting operation acting on the candidate associated data, and generating adjusting associated data according to the data adjusting operation.

In one embodiment, when the service system to be analyzed is a medical insurance fund service system, the plurality of subsystem models include a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relationship model, and data processing is performed on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result, including:

performing data processing on data associated with the medical insurance resource storage model in the input data set through the medical insurance resource storage model to obtain a medical insurance resource storage result; data processing is carried out on data which are associated with the medical insurance service relation model in the input data set through the medical insurance service relation model, and a medical insurance service relation result is obtained; performing data processing on data associated with the medical insurance resource consumption model, a medical insurance resource storage result and a medical insurance service relation result in the input data set through the medical insurance resource consumption model to obtain a medical insurance resource consumption result; and obtaining a service simulation result based on the medical insurance resource storage result, the medical insurance service relation result and the medical insurance resource consumption result.

A data processing apparatus based on a system dynamics model, the apparatus comprising:

the data acquisition module is used for acquiring an input data set corresponding to the service system to be analyzed; the input data set comprises prediction related data and adjustment related data;

the data input module is used for inputting the input data set into a system dynamic model corresponding to the service system to be analyzed; the system dynamics model comprises a plurality of subsystem models with system dynamics connections; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of prediction associated data and adjustment associated data;

the data processing module is used for respectively carrying out data processing on data related to the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result; the service simulation result comprises a data processing result obtained by processing data by different subsystem models.

In one embodiment, the data acquisition module is further configured to display a resource analysis interface corresponding to the service system to be analyzed; displaying candidate associated data on the resource analysis interface; and receiving data adjusting operation acting on the candidate associated data, and generating adjusting associated data according to the data adjusting operation.

In one embodiment, the predicted associated data includes associated data of at least two data types, and the data obtaining module is further configured to obtain at least two historical associated data corresponding to each data type; the historical associated data carries time information; forming historical associated data sequences by using the historical associated data corresponding to the same data type according to the time information to obtain the historical associated data sequences corresponding to various data types; acquiring data prediction models corresponding to various data types; inputting each historical associated data sequence into a corresponding data prediction model to obtain prediction results corresponding to various data types; the individual prediction results constitute prediction-related data.

In one embodiment, the data obtaining module is further configured to determine, in the current historical associated data sequence, a processing priority of each historical associated data based on an arrangement order of each historical associated data; acquiring a reference state parameter corresponding to the current data type; performing feature extraction on the historical associated data of the current processing priority based on the reference state parameters to obtain feature extraction results, taking the feature extraction results as the reference state parameters in the next round of feature extraction, and returning to the step of performing feature extraction on the historical associated data of the current processing priority based on the reference state parameters until the feature extraction of each historical associated data is completed to obtain target feature extraction results; and performing fusion processing on the target feature extraction result to obtain a prediction result corresponding to the current data type.

In one embodiment, the reference state parameters include a first state parameter and a second state parameter, and the data acquisition module is further configured to perform data splicing on the first state parameter and the historical associated data of the current processing priority to obtain spliced data; acquiring a forgetting parameter, and forgetting the spliced data based on the forgetting parameter to obtain a forgetting factor; acquiring an updating parameter, and updating the spliced data based on the updating parameter to obtain a target updating factor; performing state updating on the second state parameter based on the forgetting factor and the target updating factor to obtain a second state updating parameter; acquiring a prediction parameter, and performing prediction processing on the splicing data based on the prediction parameter to obtain a prediction factor; fusing the prediction factor and the second state updating parameter to obtain a first state updating parameter; and obtaining a feature extraction result based on the first state updating parameter and the second state updating parameter.

In one embodiment, the forgetting parameters include a forgetting matrix and a forgetting constant, and the data acquisition module is further configured to perform fusion processing on the forgetting matrix and the spliced data to obtain a forgetting fusion result; and obtaining a forgetting factor based on the forgetting constant and the forgetting fusion result.

In one embodiment, the update parameters include a first update matrix, a first update constant, a second update matrix, and a second update constant, and the data acquisition module is further configured to perform fusion processing on the first update matrix and the spliced data to obtain a first update fusion result; obtaining a first updating factor based on the first updating constant and the first updating fusion result; fusing the second updating matrix and the splicing data to obtain a second updating fusion result; obtaining a second updating factor based on the second updating constant and the second updating fusion result; and obtaining a target updating factor based on the first updating factor and the second updating factor.

In one embodiment, the target update factor includes a first update factor and a second update factor, and the data acquisition module is further configured to perform fusion processing on the forgetting factor and the second state parameter to obtain a second state forgetting result; fusing the first updating factor and the second updating factor to obtain a fused updating factor; and obtaining a second state updating parameter based on the second state forgetting result and the fusion updating factor.

In one embodiment, the prediction parameters include a prediction matrix and a prediction constant, and the data acquisition module is further configured to perform fusion processing on the prediction matrix and the splicing data to obtain a prediction fusion result; and obtaining a prediction factor based on the prediction constant and the prediction fusion result.

In one embodiment, the data obtaining module is further configured to obtain a training sample of a data prediction model to be currently trained; the training sample comprises a standard correlation data sequence and a corresponding standard prediction result; inputting the standard associated data sequence into a data prediction model to be trained at present to obtain an initial prediction result; calculating to obtain a training loss value according to the initial prediction result and the standard prediction result; and adjusting model parameters of the current data prediction model to be trained based on the training loss value until a convergence condition is met, so as to obtain the trained data prediction model.

In one embodiment, the data processing module is further configured to perform data processing on data associated with the resource reservation model in the input data set through the resource reservation model to obtain a resource reservation result; performing data processing on data associated with the resource service relationship model in the input data set through the resource service relationship model to obtain a resource service relationship result; performing data processing on data associated with the resource consumption model, a resource storage result and a resource service relation result in the input data set through the resource consumption model to obtain a resource consumption result; and obtaining a service simulation result based on the resource reservation result, the resource service relation result and the resource consumption result.

In one embodiment, the data processing module is further configured to determine that the service simulation result is a forward result when a difference between the resource reservation result and the resource consumption result is smaller than a preset threshold and a change trend of the resource service relationship result satisfies a preset condition; the change trend is determined based on the resource service relationship result and the historical service relationship result; otherwise, determining that the service simulation result is a negative result.

In one embodiment, when the service system to be analyzed is a medical insurance service system, the plurality of subsystem models include a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relationship model, and the data processing module is further configured to perform data processing on data associated with the medical insurance resource storage model in the input data set through the medical insurance resource storage model to obtain a medical insurance resource storage result; data processing is carried out on data which are associated with the medical insurance service relation model in the input data set through the medical insurance service relation model, and a medical insurance service relation result is obtained; performing data processing on data associated with the medical insurance resource consumption model, a medical insurance resource storage result and a medical insurance service relation result in the input data set through the medical insurance resource consumption model to obtain a medical insurance resource consumption result; and obtaining a service simulation result based on the medical insurance resource storage result, the medical insurance service relation result and the medical insurance resource consumption result.

In one embodiment, the data processing module is further configured to take the adjustment associated data as target adjustment associated data when the service simulation result is a forward result; and when the service simulation result is a negative result, updating the adjustment associated data, inputting the prediction associated data and the updated adjustment associated data into the system dynamic model until the updated service simulation result is a positive result, and taking the corresponding adjustment associated data as target adjustment associated data.

A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

The data processing method, the device, the computer equipment and the storage medium based on the system dynamics model are characterized in that an input data set corresponding to a service system to be analyzed is obtained, the input data set comprises prediction associated data and adjustment associated data, the input data set is input into a system dynamics model corresponding to the service system to be analyzed, the system dynamics model comprises a plurality of subsystem models with system power connection, the subsystem models comprise a resource storage model, a resource consumption model and a resource service relation model, the resource storage model, the resource consumption model and the resource service relation model are a causal relation network established based on the causal relation of the prediction associated data and the adjustment associated data, and the data related to the subsystem models in the input data set are respectively processed through different subsystem models in the system dynamics model, and obtaining a service simulation result, wherein the service simulation result comprises a data processing result obtained by processing data by different subsystem models. Therefore, a system dynamics model for big data analysis is established in advance based on the causal relationship among the associated data, the system dynamics model comprehensively considers the influence among a large amount of associated data, complex data processing can be carried out, and an accurate data processing result is obtained. When the global service result of the service system is predicted, the system dynamics model can output an accurate service simulation result only by inputting the predicted associated data and the adjusted associated data into the system dynamics model. Furthermore, the system dynamics model comprises a plurality of subsystem models with system dynamics connection, and different subsystem models can synchronously carry out data processing on corresponding input data, so that the data processing efficiency can be improved.

Drawings

FIG. 1 is a diagram of an exemplary system dynamics model-based data processing method;

FIG. 2 is a schematic flow chart diagram of a data processing method based on a system dynamics model in one embodiment;

FIG. 3 is a schematic flow chart illustrating data processing performed by the data prediction model in one embodiment;

FIG. 4 is a flowchart illustrating feature extraction performed on historical associated data of a current processing priority based on a reference state parameter to obtain a feature extraction result in one embodiment;

FIG. 5 is a schematic diagram of a process for memory cell based feature extraction in one embodiment;

FIG. 6 is a schematic diagram of another embodiment of a process for memory cell based feature extraction;

FIG. 7 is a schematic flow chart illustrating a service simulation result obtained based on a system dynamics model in one embodiment;

FIG. 8A is a schematic diagram of a medical insurance resource reservation model in an embodiment;

FIG. 8B is a diagram illustrating a structure of a relationship model for a town medical insurance service in an embodiment;

FIG. 8C is a schematic diagram of a structure of the city house medical insurance service relationship model in one embodiment;

FIG. 8D is a diagram illustrating an exemplary architecture of a medical insurance resource consumption model;

FIG. 9A is a schematic diagram of an interface of the medical insurance service system in one embodiment;

FIG. 9B is a schematic diagram of an interface of the medical insurance service system in another embodiment;

FIG. 9C is a schematic diagram of an interface of a medical insurance service system in yet another embodiment;

FIG. 9D is a schematic flow chart illustrating a data processing method based on a system dynamics model according to another embodiment;

FIG. 10 is a block diagram of a data processing apparatus based on a system dynamics model according to an embodiment;

FIG. 11 is a diagram of the internal structure of a computer device in one embodiment;

FIG. 12 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The data processing method based on the system dynamics model can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers.

The terminal 102 and the server 104 can be used separately to execute the data processing method based on the system dynamics model provided in the embodiment of the present application. For example, the server 104 obtains an input data set corresponding to the service system to be analyzed, where the input data set includes the prediction related data and the adjustment related data. The server 104 inputs the input data set into a system dynamics model corresponding to the service system to be analyzed, wherein the system dynamics model includes a plurality of subsystem models having system power connection, the plurality of subsystem models includes a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are a causal relationship network established based on causal relationships of prediction associated data and adjustment associated data. The server 104 performs data processing on the data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model, respectively, to obtain a service simulation result, where the service simulation result includes a data processing result obtained by performing data processing on the different subsystem models.

The terminal 102 and the server 104 may also be cooperatively used to perform the network access analysis method provided in the embodiments of the present application. For example, the server 104 obtains an input data set corresponding to the service system to be analyzed from the terminal 102, where the input data set includes the prediction related data and the adjustment related data. The server 104 inputs the input data set into a system dynamics model corresponding to the service system to be analyzed, wherein the system dynamics model includes a plurality of subsystem models having system power connection, the plurality of subsystem models includes a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are a causal relationship network established based on causal relationships of prediction associated data and adjustment associated data. The server 104 performs data processing on the data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model, respectively, to obtain a service simulation result, where the service simulation result includes a data processing result obtained by performing data processing on the different subsystem models.

Artificial Intelligence (AI) is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human Intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.

With the research and progress of artificial intelligence technology, the artificial intelligence technology is developed and applied in a plurality of fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, unmanned aerial vehicles, robots, smart medical care, smart customer service, and the like.

The scheme provided by the embodiment of the application relates to the technologies of machine learning, big data processing and the like of artificial intelligence, and is specifically explained by the following embodiment.

In one embodiment, as shown in fig. 2, a data processing method based on a system dynamics model is provided, which is described by taking the method as an example applied to a computer device in fig. 1, where the computer device may be the terminal 102 or the server 104 in fig. 1. Referring to fig. 2, the data processing based on the system dynamics model includes the following steps:

step S202, acquiring an input data set corresponding to a service system to be analyzed; the input data set includes prediction correlation data and adjustment correlation data.

The service system to be analyzed refers to a service system to be subjected to data analysis. The service system refers to a system for providing data analysis services for local governments, enterprises, users, and the like. The data analysis service may specifically include a data query service, a data adjustment service, a data prediction service, a data storage service, a data early warning service, and the like. The data query service is used for querying data, for example, a user can query the service system for a historical value of any associated data. The adjustment service is used for adjusting data, and can perform adjustment processing such as inputting, adding, deleting, reducing and the like on the data. The data prediction service is used to predict future data, e.g., to predict future consumption of resources, and to predict the value of any associated data over a future time period. The data storage service is used to store data, for example, resource consumption results when resource consumption results are predicted that characterize future consumption of resources. The data early warning service is used for sending early warning information according to the possibility precursor obtained through data prediction, data analysis and data processing before damage and problems occur, and reminding related personnel, so that the occurrence of damage is avoided, and loss is reduced to the greatest extent. For example, when the resource consumption result is determined to far exceed the resource storage result according to the data prediction result, it indicates that the resource balance is unbalanced, and at this time, the warning information may be generated.

The associated data refers to various data which can directly or indirectly affect the global service result of the service system. The association data may particularly comprise predictable association data and adjustable association data. The predicted associated data refers to associated data obtained by computer processing prediction and used for characterizing future data. The predictive relevance data may include at least one relevance data. The predicted associated data may be associated data calculated by using historical associated data and a custom formula, for example, the number of visitors of the next year is calculated by using the custom formula, specifically, the average change rate may be calculated according to the number of visitors of the resource in the past year, and the number of visitors of the next year and the average change rate are multiplied to obtain the number of visitors of the next year. The prediction related data may be related data predicted by a machine learning model, and for example, the number of resource visitors in the past year may be input to a trained machine learning model to predict the number of resource visitors in the next year. The adjustment-related data is related data determined by a user, collected by the user, and adjusted by the user. The adjustment association data may comprise at least one association data. The adjustment associated data may be specifically associated data input and adjusted on a corresponding interface by a user.

Specifically, a variety of service systems, such as a firearm resource service system, a vehicle resource service system, a medical insurance resource service system, and other resource management service systems, may be established in advance on the computer device. When the user needs to predict the influence of the corresponding associated data on the service result of the service system and needs to simulate the influence of the corresponding associated data on the service result of the service system, the user can trigger and generate an input data set corresponding to the service system to be analyzed in local or other computer equipment. After the computer equipment acquires the input data set corresponding to the service system to be analyzed, the data processing can be carried out on the input data set through the service system to be analyzed, and a service simulation result corresponding to the input data set is obtained.

In one embodiment, the custom formula for obtaining the prediction related data may be a specific formula corresponding to a gray prediction algorithm, a regression prediction algorithm, or the like. The gray prediction algorithm may be a GM (1,1) algorithm and the regression prediction algorithm may be a linear regression prediction algorithm. The machine learning model for obtaining the predictive relevance data may be a specific algorithm or a plurality of algorithms in machine learning and deep learning, such as a support vector machine, AdaBoosting, a convolutional neural network, a cyclic neural network, and the like. The Recurrent neural network includes, but is not limited to, LSTM (Long-short Term Memory), GRU (gated Recurrent Unit), Transformer, deep (autoregressive Recurrent neural network), and the like.

Step S204, inputting the input data set into a system dynamics model corresponding to the service system to be analyzed; the system dynamics model comprises a plurality of subsystem models with system dynamics connections; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of prediction correlation data and adjustment correlation data.

The system dynamics model is an algorithm model established based on system dynamics. The System Dynamics (SD) is to study a feedback System by establishing a flow level and flow rate System, define a complex System as a feedback structure with high order, multiple loops and nonlinearity, and use a causal graph and a System flow graph to represent the structure of the System. The causal graph is a causal network established based on causal relationships between data. The causal relationship between the data may be embodied by a computational expression, for example, the data a is positively influenced by the data B and the data C, and the degree of positive influence of the data B and the data C on the data a is obtained by big data analysis, and then the computational expression may be data a ═ data B × B + data C. And B represents an influence factor corresponding to the data B and represents the influence degree of the data B on the data A, wherein the positive influence is when B is larger than 0, and the negative influence is when B is smaller than 0. C represents an influence factor corresponding to the data C and represents the influence degree of the data C on the data A, wherein the positive influence is when C is larger than 0, and the negative influence is when C is smaller than 0. The system flow graph is a system flow direction network established based on the incidence relation between causal relationship networks. Associations between causal networks may be determined based on causal relationships between association data in one causal network and association data in another causal network.

The system dynamics model includes a plurality of subsystem models having system dynamics associations. The system dynamics model is an integral model. It can be understood that the system dynamics model has a plurality of functions, a plurality of data processing results can be obtained through data processing, and one function corresponds to at least one data processing result. The subsystem model is a sub-model composed of associated data corresponding to a specific function, and is a part of the system dynamics model. And the subsystem models form a system dynamic model. And each subsystem model has system power connection, and can carry out data interaction based on the system power connection, so that the final data processing result is influenced. For example, the data processing result or the intermediate processing result of one subsystem model may flow into another subsystem model, and affect the data processing result of another subsystem model together with the input data of another subsystem model. Wherein the system power connection is determined based on causal relationships between the associated data.

The system dynamics model can specifically have a resource reservation statistical function, a resource consumption statistical function and a resource service relationship statistical function. The resources refer to social resources, and may specifically include human resources, information resources, and various material resources (e.g., material wealth) created through labor. The resource reserve model is a sub-model composed of associated data corresponding to a resource reserve statistical function. Resource reservation refers to the reservation action taken on a resource during the resource flow. The resource reserve statistics refers to statistics of reserve amounts generated by various reserve behaviors, and resource reserve statistics results are obtained based on the various reserve amounts. For example, during the resource flow process, the total reserve amount of the resource is counted. The method specifically includes obtaining associated data which can affect the resource reserve statistical result from all associated data, that is, obtaining associated data corresponding to the resource reserve statistical function, and establishing a resource reserve model based on a causal relationship between the obtained associated data and the associated data.

The resource consumption model is a sub-model composed of associated data corresponding to a resource consumption statistical function. Resource consumption refers to the consumption behavior that is taken on a resource during the resource flow. The resource consumption statistics refers to statistics of consumption generated by various consumption behaviors, and a resource consumption statistical result is obtained based on various consumption. For example, in the resource flow process, the total resource consumption is counted. The method specifically includes obtaining associated data that may affect the resource consumption statistical result from all associated data, that is, obtaining associated data corresponding to the resource consumption statistical function, and establishing a resource consumption model based on a causal relationship between the obtained associated data and the associated data.

The resource service relation model is a sub-model consisting of associated data corresponding to the resource service relation statistical function. A resource service relationship refers to a service relationship between a resource service provider and a resource consumer. In the resource consumption process, the resource consumer may specifically consume the resource through the resource consumption service provided by the resource service provider. For example, a police officer arriving at a police team to pick up a firearm may consume firearm resources. The resource service relationship statistics refers to the statistics of the providing capability of the resource consumption service of each resource service provider to obtain the resource service relationship statistical result. For example, in the process of firearm resource consumption, firearm banks and ammunition banks of various police teams are counted. And in the process of medical insurance resource consumption, the outpatient service volume and the hospitalization volume of each stage of medical institution are counted. The method specifically includes obtaining associated data which may affect the resource service relationship statistical result from all associated data, that is, obtaining associated data corresponding to the resource service relationship statistical function, and establishing a resource service relationship model based on a causal relationship between the obtained associated data and the associated data. The resource service relationship statistical result can reflect the resource service effect.

The associated data sets corresponding to the resource reservation model, the resource consumption model and the resource service relationship model comprise the same associated data and different associated data. That is, some associated data may affect the resource reservation statistics result, the resource consumption statistics result, and the resource service relationship statistics result, and some data may only affect the resource consumption statistics result. It is understood that there is an overlapping model structure between the resource reservation model, the resource consumption model, and the resource service relationship model.

Furthermore, different regions can establish different system dynamics models according to the characteristics of the regions.

Step S206, respectively carrying out data processing on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result; the service simulation result comprises a data processing result obtained by processing data by different subsystem models.

The service simulation result refers to a simulation result obtained after the system dynamics model performs data processing on the input data set, and is used for reflecting a global service influence result generated by the service system based on the input data set. The service simulation result may include a data processing result obtained by performing data processing on each subsystem model, and may further include a service influence result, where the service influence result includes a positive result and a negative result. A positive result indicates that the global service impact result generated based on the input data set is positive, contributing. A negative result indicates that the global service impact result generated based on the input data set is negative, hindering.

Specifically, after the computer device inputs the input data set corresponding to the service system to be analyzed into the system dynamic model corresponding to the service system to be analyzed, each input data performs data fusion and data flow according to the data flow direction and the calculation expression in the system dynamic model, and outputs the final data processing result. Further, the system dynamics model includes a resource reservation model, a resource consumption model and a resource service relationship model with system dynamics relation, the resource reservation model, the resource consumption model and the resource service relationship model can respectively obtain corresponding input data from the input data set for data processing to obtain corresponding data processing results, and the service simulation results are generated based on the data processing results respectively obtained by the resource reservation model, the resource consumption model and the resource service relationship model. After obtaining the service simulation result, the computer device may display the service simulation result on a related interface of the local or other computer device. Further, when the service simulation result is a positive result, the computer device may store and recommend the corresponding input data set, and when the service simulation result is a negative result, the computer device may perform early warning.

For example, the medical insurance service system is a complex system influenced by a plurality of factors such as population, economy, national policy and the like, the main bodies comprise governments, medical insurance departments, medical institutions, insurers and the like, and the interaction among the factors can influence the medical insurance service. By collecting and analyzing data such as policy documents, reference documents, past research and the like related to medical insurance resource storage, medical insurance resource consumption and medical insurance service relationship, related influence factors can be determined, so that medical insurance related data can be determined, and the causal relationship among the related influence factors can be determined, so that the causal relationship among the medical insurance related data can be determined. And establishing a system dynamic model corresponding to the medical insurance service system based on the causal relationship among the medical insurance associated data. The system dynamics model corresponding to the medical insurance service system comprises a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relation model with system dynamic connection. Inputting the medical insurance related data of the past year into a corresponding machine learning model to predict part of medical insurance related data of the next year, namely predicted related data, and acquiring part of medical insurance related data of the next year, namely adjusted related data, which is obtained through adjustment and collection by a user. And the predicted associated data and the adjusted associated data form an input data set, and the computer equipment inputs the input data set into a system dynamic model corresponding to the medical insurance service system. When the system dynamics model processes the input data set, specifically, data associated with the medical insurance resource storage model is input into the medical insurance resource storage model to obtain a medical insurance resource storage result, data associated with the medical insurance resource consumption model is input into the medical insurance resource consumption model to obtain a medical insurance resource consumption result, and data associated with the medical insurance service relationship model is input into the medical insurance service relationship model to obtain a medical insurance service relationship result. And obtaining a medical insurance service simulation result based on the medical insurance resource storage result, the medical insurance resource consumption result and the medical insurance service relation result. The system dynamics model corresponding to the medical insurance service system can output the medical insurance resource storage result, the medical insurance resource consumption result, the medical insurance service relation result and the medical insurance service influence result, and the computer equipment can display the output result.

In the data processing method based on the system dynamics model, the input data set corresponding to the service system to be analyzed is obtained and comprises the prediction correlation data and the adjustment correlation data, the input data set is input into the system dynamics model corresponding to the service system to be analyzed, the system dynamics model comprises a plurality of subsystem models with system power connection, the subsystem models comprise a resource storage model, a resource consumption model and a resource service relation model, the resource storage model, the resource consumption model and the resource service relation model are causal relationship networks established based on the causal relationship of the prediction correlation data and the adjustment correlation data, and the data related to the subsystem models in the input data set are respectively processed through different subsystem models in the system dynamics model to obtain the service simulation result, the service simulation result comprises a data processing result obtained by processing data by different subsystem models. Therefore, a system dynamics model for big data analysis is established in advance based on the causal relationship among the associated data, the system dynamics model comprehensively considers the influence among a large amount of associated data, complex data processing can be carried out, and an accurate data processing result is obtained. When the global service result of the service system is predicted, the system dynamics model can output an accurate service simulation result only by inputting the predicted associated data and the adjusted associated data into the system dynamics model. Furthermore, the system dynamics model comprises a plurality of subsystem models with system dynamics connection, and different subsystem models can synchronously carry out data processing on corresponding input data, so that the data processing efficiency can be improved.

In one embodiment, obtaining an input data set corresponding to a service system to be analyzed includes: displaying a resource analysis interface corresponding to the service system to be analyzed; displaying candidate associated data on the resource analysis interface; and receiving data adjusting operation acting on the candidate associated data, and generating adjusting associated data according to the data adjusting operation.

The resource analysis interface is a human-computer interaction interface corresponding to the service system and can be used for adjusting the associated data, displaying the associated data and displaying the data processing result. In one embodiment, the resource analysis interface may include a data adjustment area and a data presentation area. The data adjusting area is used for adjusting the associated data. And the data display area is used for displaying the data processing result of the system dynamics model and displaying the prediction correlation data. The candidate associated data refers to adjustable associated data. Data adjustment operations include, but are not limited to, selection operations, input operations, sliding operations, and the like. Input operations include, but are not limited to, text input, voice input, and the like.

Specifically, the computer device may display a resource analysis interface corresponding to the service system to be analyzed to the user, where candidate associated data that can be adjusted by the user is displayed on the resource analysis interface. The display mode of the candidate associated data may be that a default value or a historical value of the candidate associated data and an adjustment control for adjusting the default value or the historical value are displayed on the resource analysis interface, so that the user may slide the adjustment control to adjust the default value or the historical value to obtain the adjustment associated data. Correspondingly, the computer equipment can receive the sliding operation acted on the adjusting control by the user and generate the adjusting associated data according to the sliding operation. The display mode of the candidate associated data may also be to display an input box corresponding to the candidate associated data on the resource analysis interface, so that the user may input a specific numerical value in the input box to obtain the adjusted associated data. Accordingly, the computer device can receive the input operation of the user on the input box, and generate the adjustment associated data according to the input operation.

In the embodiment, the candidate associated data is displayed through the visual interface for the user to adjust, and the method is convenient and quick.

In one embodiment, the predicting the associated data includes associated data of at least two data types, and acquiring an input data set corresponding to the service system to be analyzed includes: acquiring at least two historical associated data respectively corresponding to various data types; the historical associated data carries time information; forming historical associated data sequences by using the historical associated data corresponding to the same data type according to the time information to obtain the historical associated data sequences corresponding to various data types; acquiring data prediction models corresponding to various data types; inputting each historical associated data sequence into a corresponding data prediction model to obtain prediction results corresponding to various data types; the individual prediction results constitute prediction-related data.

The historical associated data refers to associated data generated in a historical time period, such as the number of resource visitors in the last year. The time information refers to the historical time corresponding to the historical associated data. The data prediction model is a machine learning model for predicting future associated data. In order to ensure the accuracy of data prediction, the associated data of one data type corresponds to a data prediction model, for example, the data prediction model is trained to be specially used for predicting the number of the future resource visitors, and the data prediction model is trained to be specially used for predicting the number of the future per-person resource visitors.

Specifically, for associated data having a time dependency relationship and capable of changing according to a long-term trend, future data can be predicted by a machine learning model. It is understood that the system dynamics model relates to a large amount of associated data, and typically, the associated data that can be predicted by the machine learning model includes associated data of at least two data types, and the associated data of each data type needs to be subjected to data prediction by a corresponding data prediction model. When predicting data, the computer device may obtain at least two historical associated data corresponding to time corresponding to each data type, and sort the historical associated data corresponding to the same data type according to the sequence of time from the morning to the evening to form a historical associated data sequence. The historical associated data sequence is input into a data prediction model, the data prediction model can sequentially learn the characteristic information of each historical associated data according to the time sequence, and the future associated data can be predicted according to the learned characteristic information. Since there are at least two data types of associated data, at least two historical associated data sequences can be obtained. Further, the computer device obtains data prediction models corresponding to various data types, inputs various historical associated data sequences into the corresponding data prediction models to obtain prediction results corresponding to various data types, and finally, the prediction results form prediction associated data.

For example, the forecast associated data includes the number of resource visits and the number of per-capita resource visits in the next year. The computer equipment can obtain the historical resource access people number sequence of 'the resource access people number in the previous year, the resource access people number in the last year and the resource access people number in the present year', and the historical resource access people number sequence is input into a data prediction model for predicting the resource access people number to obtain the resource access people number in the next year. The computer equipment can obtain the number of times of visiting per capita resources of the previous year, the last year and the present year to form a historical number of times of visiting per capita resources of the previous year, the number of times of visiting per capita resources of the last year and the number of times of visiting per capita resources of the present year, and the historical number of times of visiting per capita resources is input into a data prediction model for predicting the number of times of visiting per capita resources of the present year to obtain the number of times of visiting per capita resources of the present year.

In the embodiment, the associated data of different data types are predicted based on different data prediction models, so that the accuracy of data prediction can be improved.

In one embodiment, as shown in fig. 3, inputting each historical associated data sequence into a corresponding data prediction model to obtain a prediction result corresponding to each data type, including:

step S302, in the current history related data sequence, the processing priority of each history related data is determined based on the arrangement sequence of each history related data.

The current historical associated data sequence refers to a current historical associated data sequence to be processed. The processing priority refers to the processing sequence of the data prediction model for processing each historical associated data in the historical associated data sequence. The larger the time difference between the time information of the historical associated data and the current time is, the higher the corresponding processing priority is, that is, the earlier the time corresponding to the historical associated data is, the higher the corresponding processing priority is. For example, the processing priority corresponding to the historical associated data of the previous year is higher than the processing priority corresponding to the historical associated data of the last year.

Specifically, one history associated data sequence may be randomly determined from the respective history associated data sequences as the current history associated data sequence. In the current historical associated data sequence, the computer device may determine a processing priority for each historical associated data based on the rank order of each historical associated data. If the historical associated data in the current historical associated data sequence are sorted from morning to evening, the processing priority corresponding to the historical associated data which is sorted more forward is higher, and if the historical associated data in the current historical associated data sequence are sorted from evening to morning, the processing priority corresponding to the historical associated data which is sorted more backward is higher.

Step S304, obtaining the reference state parameter corresponding to the current data type.

And S306, extracting the features of the historical associated data of the current processing priority based on the reference state parameters to obtain a feature extraction result, taking the feature extraction result as the reference state parameters in the next round of feature extraction, and returning to the step of extracting the features of the historical associated data of the current processing priority based on the reference state parameters until the feature extraction of each historical associated data is completed to obtain a target feature extraction result.

The data prediction model comprises an input layer, a hidden layer and an output layer, wherein the input layer is connected with the hidden layer, and the hidden layer is connected with the output layer. The input layer comprises at least two input nodes, and one input node corresponds to one historical associated data. The input layer may convert the historical associated data into vectors for subsequent data processing. The hidden layer comprises at least one hidden node, each hidden node comprises at least two memory cells, and each hidden node is used for extracting characteristic information of different dimensions of the historical associated data. And each memory cell in the same hidden node is used for processing historical associated data corresponding to different times, and the memory cells are sequentially connected according to a time sequence. The memory cell corresponding to the starting time can learn the characteristic information of the historical associated data corresponding to the starting time based on the starting state parameter, the learned characteristic information is transmitted to the memory cell corresponding to the next time, the memory cell can carry out combined learning on the transmitted characteristic information and the corresponding historical associated data, the learned new characteristic information is further transmitted to the memory cell corresponding to the next time, and by analogy, the target characteristic information is obtained after all the memory cells finish learning. The reference state parameter refers to a starting state parameter. The reference state parameter corresponding to the current data type refers to an initial state parameter in the current data prediction model. Different data prediction models correspond to different reference state parameters. The feature extraction result refers to feature information extracted and learned by the memory cells from the received data. The target feature extraction result refers to the feature information finally learned, and the target feature extraction result is a fusion result of the feature information learned by each memory cell. The output layer can perform fusion processing on the target feature extraction result and finally output a prediction result.

Specifically, when data prediction is performed, the data prediction models used by the computer device are trained data prediction models. The trained data prediction model comprises model parameters obtained by training based on a large number of training samples, and the reference state parameters are part of the model parameters. The computer equipment can obtain a reference state parameter corresponding to a current data prediction model, namely, the reference state parameter corresponding to a current data type, perform feature extraction on historical associated data with a first processing priority based on the reference state parameter to obtain a first feature extraction result, perform feature extraction on historical associated data with a second processing priority based on the first feature extraction result to obtain a second feature extraction result, perform feature extraction on historical associated data with a third processing priority based on the second feature extraction result to obtain a third feature extraction result, and so on until all historical associated data are subjected to feature extraction, and finally obtain a target feature extraction result.

In one embodiment, the calculation formula of the number of hidden nodes of the hidden layer is as follows:

where N represents the number of hidden nodes, N represents the number of input nodes, m represents the number of output nodes, and a is a constant, typically a constant of 1-10.

And step S308, fusing the target feature extraction results to obtain a prediction result corresponding to the current data type.

Specifically, the computer device may perform fusion processing on the target feature extraction result to obtain a prediction result corresponding to the current data type. The data processing of the hidden layer is the operation between vectors, the target feature extraction result output by the hidden layer is a multi-dimensional vector, the output layer needs to perform fusion processing on the multi-dimensional vector, the multi-dimensional features are fused together, one-dimensional data is output as a prediction result, and the prediction result with a specific numerical value is output. For example, a data prediction model for predicting resource visitors ultimately outputs the next year of resource visitors.

In one embodiment, the output layer may be an MLP (Multi-layer Perceptron) layer.

In one embodiment, when the influence factors of the associated data to be predicted are more, in addition to predicting the predicted associated data corresponding to the data type based on a plurality of historical associated data corresponding to the data type of the associated data to be predicted, the predicted associated data corresponding to the data type can be predicted together with the influence factors of the associated data to be predicted, so that the data prediction accuracy is improved. For example, the input data of the data prediction model for predicting the per-capita GDP may be a historical per-capita GDP sequence, or a historical related data sequence composed of GDP influencing factors such as national revenue, national foreign exchange reserve, currency supply, gold reserve, total release capital, factory price index of industrial products, non-manufacturing procurement manager index and the like over the years. GDP influence factors such as national revenue, national foreign exchange reserve, currency supply, gold reserve, total issued capital, industrial product delivery price index, non-manufacturing purchasing manager index and the like form a vector corresponding to historical associated data.

In one embodiment, the number of output nodes of the output layer may determine the amount of prediction data output by the data prediction model. For example, the number of output nodes of the output layer is 2, and the data prediction model can output prediction related data of next year and later year. And obtaining the forecast related data of the next year based on the historical related data sequence and the forecast related data of the next year. That is, after the data prediction model obtains the next year of prediction related data, the prediction related data can be input into the hidden layer, and the hidden layer can further perform data processing on the next year of prediction related data on the basis of the processing result of the historical related data sequence to obtain the next year of prediction related data. The method for processing the forecast related data in the next year is the same as the method for processing the historical related data.

In the embodiment, the characteristic information of each historical associated data is sequentially extracted according to the time sequence, so that the complex rule in the time sequence can be learned, and the time dependence relation existing among the historical associated data can be learned, thereby accurately predicting the future associated data and improving the accuracy of data prediction.

In one embodiment, as shown in fig. 4, the reference state parameters include a first state parameter and a second state parameter, and feature extraction is performed on the historical associated data of the current processing priority based on the reference state parameters to obtain a feature extraction result, including:

and step S402, performing data splicing on the first state parameter and the historical associated data of the current processing priority to obtain spliced data.

The first state parameter is used for determining data needing to be discarded and forgotten from the historical associated data, and therefore data needing to be reserved and added are extracted. The second state parameter is used for fusing the related data of the previous historical related data into the current historical related data.

Specifically, the computer device may perform data splicing on the first state parameter and the historical associated data of the current processing priority to obtain spliced data. The first state parameter and the historical associated data of the current processing priority are vectors, and the data splicing specifically includes left-right splicing, up-down splicing and the like of the two vectors.

And S404, acquiring a forgetting parameter, and forgetting the spliced data based on the forgetting parameter to obtain a forgetting factor.

Wherein, the forgetting parameter is used for forgetting the data. The forgetting factor obtained by the forgetting process is used to control which information of the second state parameter and the current history related data is allowed to enter the current memory cell, that is, which information of the last memory cell and the current history related data needs to be forgotten.

Specifically, the computer device can obtain a forgetting parameter, forget the spliced data based on the forgetting parameter, determine information to be discarded and forgotten in the spliced data, and generate a forgetting factor. The forgetting processing can be specifically a forgetting formula designed for forgetting processing, variables of the forgetting formula comprise a forgetting parameter and splicing data, and the forgetting factor can be obtained by inputting the forgetting parameter and the splicing data into the forgetting formula.

In one embodiment, the forgetting parameters include a forgetting matrix and a forgetting constant, and forgetting the spliced data based on the forgetting parameters to obtain a forgetting factor, including: fusing the forgetting matrix and the splicing data to obtain a forgetting fusion result; and obtaining a forgetting factor based on the forgetting constant and the forgetting fusion result.

Specifically, the forgetting parameters include a forgetting matrix and a forgetting constant, the forgetting matrix is used for performing weighted calculation on the spliced data, and the forgetting constant is used for constraining a weighted calculation result. The computer equipment can perform fusion processing on the forgetting matrix and the splicing data to obtain a forgetting fusion result. The fusion process may specifically be a process such as matrix multiplication or matrix addition. After the forgetting fusion result is obtained, the computer device may obtain a forgetting factor based on the forgetting constant and the forgetting fusion result, specifically, may calculate a statistical value of the forgetting constant and the forgetting fusion result, and map the statistical value into a preset numerical range to obtain the forgetting factor. The statistical value for calculating the forgetting constant and the forgetting fusion result may specifically be a sum, a weighted sum, an average value, a weighted average value, and the like of the calculated forgetting constant and the forgetting fusion result. The mapping of the statistical value to the preset value range may specifically be mapping the statistical value to the preset value range through a preset function. For example, the statistics may be mapped by a Sigmoid activation function between 0 and 1, with 0 indicating total forgetting and 1 indicating no forgetting, i.e., a forgetting factor between 0 and 1. The statistics can also be mapped between-1 and 1 by the tanh activation function, with negative numbers indicating reverse forgetting and positive numbers indicating forward forgetting, i.e., a forgetting factor between-1 and 1. The preset numerical range can be a numerical range set according to actual needs.

In one embodiment, the calculation formula for forgetting processing is as follows: f. of_t＝σ(W_f·[h_t-1，x_t]+b_f). Wherein f is_tRepresenting a forgetting factor,. sigma.representing a Sigmoid activation function, W_fRepresenting a forgetting matrix, i.e. a forgetting weight, h_t-1Representing a first state parameter, x_tHistory association data representing the priority of the current processing]Represents the concatenation of vectors, [ h ]_t-1，x_t]Representing spliced data, b_fIndicating a forgetting constant.

And step S406, acquiring an updating parameter, and updating the spliced data based on the updating parameter to obtain a target updating factor.

The updating parameter is used for updating the data. The target update factor obtained by the update process is used to control which information of the first state parameter and the current historical association data is allowed to enter the current memory cell. The forgetting factor and the target update factor can jointly determine which information of the last memory cell is allowed to enter the current memory cell.

Specifically, the computer device may obtain the update parameter, perform update processing on the splicing data based on the update parameter, determine information to be updated in the splicing data, and generate the target update factor. The updating process may specifically be designing an updating formula for the updating process, where variables of the updating formula include an updating parameter and splicing data, and the updating parameter and the splicing data are input into the updating formula to calculate a target updating factor.

In one embodiment, the updating parameters include a first updating matrix, a first updating constant, a second updating matrix and a second updating constant, and the updating process is performed on the spliced data based on the updating parameters to obtain the target updating factor, including: fusing the first updating matrix and the splicing data to obtain a first updating fusion result; obtaining a first updating factor based on the first updating constant and the first updating fusion result; fusing the second updating matrix and the splicing data to obtain a second updating fusion result; obtaining a second updating factor based on the second updating constant and the second updating fusion result; and obtaining a target updating factor based on the first updating factor and the second updating factor.

Specifically, the update parameters include a first update matrix, a first update constant, a second update matrix, and a second update constant, where the update matrix is used to perform weighted calculation on the spliced data, and the update constant is used to constrain the weighted calculation result. The first update matrix and the second update matrix are different update matrices, and can perform weighted calculation of different weights on the spliced data, and the first update constant and the second update constant are different update constants, and can perform constraint of different degrees on corresponding weighted calculation results. The computer device can perform fusion processing on the first update matrix and the splicing data to obtain a first update fusion result. The fusion process may specifically be a process such as matrix multiplication or matrix addition. After obtaining the first update fusion result, the computer device may obtain a first update factor based on the first update constant and the first update fusion result, specifically, may calculate a statistical value of the first update constant and the first update fusion result, and map the statistical value into a preset numerical range to obtain the first update factor. The computer device may perform fusion processing on the second update matrix and the splicing data to obtain a second update fusion result, obtain a second update factor based on the second update constant and the second update fusion result, and finally form a target update factor by the first update factor and the second update factor. The statistics of calculating the updated constant and the updated fusion result may specifically be a sum, a weighted sum, an average, a weighted average, and the like of calculating the updated constant and the updated fusion result. The mapping of the statistical value to the preset value range may specifically be mapping the statistical value to the preset value range through a preset function. For example, the statistics may be mapped by a Sigmoid activation function between 0 and 1, with 0 indicating no update and 1 indicating a full update, i.e., an update factor between 0 and 1. The statistics may also be mapped between-1 and 1 by a tanh activation function, negative numbers indicating reverse updates and positive numbers indicating forward updates, i.e. an update factor between-1 and 1. The preset numerical range can be a numerical range set according to actual needs.

In one embodiment, the calculation formula for the update process is as follows: i.e. i_t＝σ(W_i·[h_t-1，x_t]+b_i)，

Wherein i_tDenotes a first update factor, W_iRepresenting a first update matrix, b_iRepresenting a first update constant.

Denotes a second update factor, W_CRepresenting a second update matrix, b_CRepresenting a second update constant. tanh represents the tanh activation function.

Step S408, updating the state of the second state parameter based on the forgetting factor and the target updating factor to obtain a second state updating parameter.

Specifically, the computer device may perform state update on the second state parameter based on the forgetting factor and the target update factor, so as to obtain a second state update parameter. The second state parameter comprises the related characteristic information of the previous historical related data, and the state updating of the second state parameter based on the forgetting factor and the target updating factor can organically combine the related characteristic information of the previous historical related data and the related characteristic information of the current historical related data to obtain the second state updating parameter. The second state updating parameter is used for fusing the related characteristic information of the previous historical related data and the related characteristic information of the current historical related data into the characteristic information of the next historical related data. The performing of the state processing may specifically be calculating a sum, a weighted sum, an average, a weighted average, and the like of the forgetting factor, the target update factor, and the second state parameter.

In one embodiment, the updating the target update factor includes a first update factor and a second update factor, and performing state update on the second state parameter based on the forgetting factor and the target update factor to obtain the second state update parameter, including: performing fusion processing on the forgetting factor and the second state parameter to obtain a second state forgetting result; fusing the first updating factor and the second updating factor to obtain a fused updating factor; and obtaining a second state updating parameter based on the second state forgetting result and the fusion updating factor.

Specifically, the computer device may perform fusion processing on the forgetting factor and the second state parameter to obtain a second state forgetting result, where the second state forgetting result is used to represent information that is allowed to enter the current memory cell in the second state parameter and the current history associated data. And the computer equipment performs fusion processing on the first updating factor and the second updating factor to obtain a fusion updating factor, wherein the fusion updating factor is used for representing the information which is allowed to enter the current memory cell in the first state parameter and the current historical associated data. And finally, the computer equipment obtains a second state updating parameter based on the second state forgetting result and the fusion updating factor, and specifically, any data such as the sum, the weighted sum, the average value, the weighted average value and the like of the second state forgetting result and the fusion updating factor is calculated as the second state updating parameter.

In one embodiment, the state update is calculated as follows:

C_trepresenting a second state update parameter, f_tDenotes a forgetting factor, C_t-1Representing a second state parameter, i_tWhich represents a first update factor that is,

representing a second update factor.

And S410, acquiring a prediction parameter, and performing prediction processing on the spliced data based on the prediction parameter to obtain a prediction factor.

Specifically, the prediction parameters are used for prediction processing of data. The prediction factor obtained by the prediction process is used to control which information of the first state parameter and the current historical associated data affects the second state parameter. The computer device can obtain the prediction parameters and perform prediction processing on the splicing data based on the prediction parameters. The prediction processing may specifically be to design a prediction formula for the prediction processing, where variables of the prediction formula include prediction parameters and splicing data, and the prediction parameters and the splicing data are input into the prediction formula to calculate to obtain a prediction factor.

In one embodiment, the prediction parameters include a prediction matrix and a prediction constant, and the prediction processing is performed on the spliced data based on the prediction parameters to obtain the prediction factors, including: fusing the prediction matrix and the splicing data to obtain a prediction fusion result; and obtaining a prediction factor based on the prediction constant and the prediction fusion result.

Specifically, the prediction parameters include a prediction matrix and a prediction constant, the prediction matrix is used for performing weighted calculation on the spliced data, and the prediction constant is used for constraining the weighted calculation result. The computer equipment can perform fusion processing on the prediction matrix and the splicing data to obtain a prediction fusion result. The fusion process may specifically be a process such as matrix multiplication or matrix addition. After the forgetting fusion result is obtained, the computer device may obtain the prediction factor based on the prediction constant and the prediction fusion result, specifically, may calculate a statistical value of the prediction constant and the prediction fusion result, and map the statistical value into a preset numerical range to obtain the prediction factor. The statistical values for calculating the predicted constant and the predicted fusion result may be, specifically, a sum, a weighted sum, an average, a weighted average, and the like of the calculated predicted constant and the predicted fusion result. The mapping of the statistical value to the preset value range may specifically be mapping the statistical value to the preset value range through a preset function. For example, the statistics may be mapped by Sigmoid activation function between 0 and 1, with 0 indicating no effect and 1 indicating a full effect, i.e. a predictor between 0 and 1. Statistics can also be mapped between-1 and 1 by the tanh activation function, negative numbers representing negative effects and positive numbers representing positive effects, i.e. the predictor is between-1 and 1. The preset numerical range can be a numerical range set according to actual needs. The forgetting parameter, the updating parameter, and the prediction parameter are all part of the model parameter.

In one embodiment, the calculation formula for the prediction process is as follows: o_t＝σ(W_o·[h_t-1，x_t]+b_o). Wherein o is_tDenotes the predictor, σ denotes the Sigmoid activation function, W_oRepresenting the prediction matrix, i.e. the prediction weights, h_t-1Representing a first state parameter, x_tHistory association data representing the priority of the current processing]Represents the concatenation of vectors, [ h ]_t-1，x_t]Representing spliced data, b_oRepresenting the predicted constant.

Step S412, the prediction factor and the second state updating parameter are fused to obtain a first state updating parameter.

Specifically, the computer device may perform fusion processing on the prediction factor and the second state update parameter to obtain the first state update parameter. The fusion process may specifically be a process such as matrix multiplication or matrix addition.

In one embodiment, the predictor and the second state update parameter are fused to obtain a calculation formula of the first state update parameter as follows: h is_t＝o_t*tanh(C_t). Wherein h is_tRepresents a first state update parameter, o_tDenotes the predictor, C_tDenotes a second state update parameter, tanh denotes tanh activates the function.

Step S514, obtaining a feature extraction result based on the first state updating parameter and the second state updating parameter.

Specifically, after the first state update parameter and the second state update parameter are obtained through the data processing, the first state update parameter and the second state update parameter constitute a feature extraction result, and the feature extraction result is used as a reference state parameter in the next round of feature extraction.

The process of feature extraction is illustrated by taking three memory cells as an example. Refer to FIG. 5, where x_t-1Representing historical associated data, x, corresponding to the last time_tIndicating historical associated data, x, corresponding to the current time_t+1Representing historical associated data corresponding to the next time. h is₀Representing a first state parameter, C, corresponding to a previous time₀Representing a second state parameter, h, corresponding to the last time_t-1Representing a first state parameter, C, corresponding to the current time_t-1A second state parameter, h, corresponding to the current time_t+1Representing a first state parameter, C, corresponding to the next time_t+1And representing a second state parameter corresponding to the next time. First memory cell based on h₀And C₀For x_t-1Performing feature extraction to obtain h_t-1And C_t-1，h_t-1And C_t-1Delivered to the next memory cell, the second memory cell based on h_t-1And C_t-1For x_tPerforming feature extraction to obtain h_tAnd C_tThe third memory cell is based on h_tAnd C_tFor x_t+1Performing feature extraction to obtain h_t+1And C_t+1，h_t+1And C_t+1To the output layer. Referring to fig. 6, a data processing process inside a memory cell is illustrated by taking a memory cell as an example. h is_t-1And x_tSpliced to obtain [ h ]_t-1，x_t]，h_t-1And x_tBy the formula one { f_t＝σ(W_f·[h_t-1，x_t]+b_f) F is obtained by calculation_t，h_t-1And x_tBy the formula two i_t＝σ(W_i·[h_t-1，x_t]+b_i) Calculating to obtain i_t，h_t-1And x_tBy the formula three

Is calculated to obtain

C_t-1、f_t、i_tAnd

by the formula four

Calculating to obtain C_t。h_t-1And x_tBy the formula five o_t＝σ(W_o·[h_t-1，x_t]+b_o) Get o by calculation_t，o_tAnd C_tBy the formula six { h_t＝o_t*tanh(C_t) H is obtained by calculation_tTo finally obtain C_tAnd h_t。

In the embodiment, the comprehensive forgetting parameter, the updating parameter, the prediction parameter and the reference state parameter can respectively extract the feature information of the historical associated data from different angles, so that abundant and accurate feature extraction results are obtained, and the accuracy of the prediction data is improved.

In one embodiment, the training process of the data prediction model is realized by the following steps; acquiring a training sample of a data prediction model to be trained currently; the training sample comprises a standard correlation data sequence and a corresponding standard prediction result; inputting the standard associated data sequence into a data prediction model to be trained at present to obtain an initial prediction result; calculating to obtain a training loss value according to the initial prediction result and the standard prediction result; and adjusting model parameters of the current data prediction model to be trained based on the training loss value until a convergence condition is met, so as to obtain the trained data prediction model.

And the standard associated data sequence and the corresponding standard prediction result are obtained based on historical associated data. For example, in 2020, 2020 and past resource visitors are known, and when a data prediction model for predicting resource visitors is trained, one of the training samples may be a standard associated data sequence consisting of "2015-2016-2017" and the standard prediction result corresponding to the standard associated data sequence is 2018.

Specifically, the training process of the data prediction model is a supervised training process. The various data prediction models correspond to different training samples. The computer equipment can obtain a training sample of a data prediction model to be trained currently, the training sample comprises a standard associated data sequence and a corresponding standard prediction result, the standard associated data sequence is input into the data prediction model to be trained currently to obtain an initial prediction result, and a training loss value is obtained through calculation according to the initial prediction result and the standard prediction result. Specifically, the training loss value can be calculated according to the initial prediction result and the standard prediction result in a preset calculation mode. For example, at least one error value of error rate, root mean square error, root mean square logarithm error and decision coefficient of the initial prediction result and the standard prediction result is calculated, and the error values are weighted and summed to obtain the training loss value. And after the training loss value is obtained through calculation, adjusting model parameters of the data prediction model according to the training loss value until a convergence condition is met, and obtaining the trained data prediction model. The convergence condition may be self-defined, for example, when the training loss value reaches a minimum value, the data prediction model may be considered to satisfy the convergence condition, so as to obtain the trained data prediction model. The model parameters can be adjusted specifically by using an Adaptive motion Estimation (Adam) optimization algorithm, so that model parameters of each layer of the data prediction model are optimized, and a final model parameter is obtained after multiple iterations, thereby obtaining a trained data prediction model. Finally, training samples corresponding to the data prediction models can be trained respectively to obtain the data prediction models capable of accurately predicting data. It is understood that, in addition to the Adam optimization Algorithm, optimization algorithms such as SGD (random Gradient Descent Algorithm), RMSprop (Root Mean Square Gradient Descent Algorithm), adarad (Adaptive Gradient Algorithm), AdaBelief (Adaptive Belief) and the like may be used.

In this embodiment, the data prediction model for predicting data can be obtained by fast training through supervised training.

In an embodiment, as shown in fig. 7, the data processing, performed by different subsystem models in the system dynamics model, on the data associated with the subsystem models in the input data set respectively to obtain the service simulation result includes:

step S702, data processing is carried out on the data which is in the input data set and is related to the resource storage model through the resource storage model, and a resource storage result is obtained.

Specifically, the input data set includes resource consumer association data, resource service provider association data, and resource service maker association data. The resource consumer-related data is data related to a resource consumer, and the resource consumer refers to a user who consumes a resource and stores the resource, for example, an employee or a resident. A resource service provider is data associated with a resource service provider. The resource service provider refers to a user who provides a resource consumption service, for example, an police team who provides a firearm picking-up service, a bookstore who provides a book ordering service, and a medical institution who provides a medical service. Resource service formulator associated data is data associated with a resource service formulator. The resource service customizer refers to a user who makes resource consumption service standards and resource storage service standards, such as police force management departments, local governments, and medical insurance management departments.

Further, the resource consumption side correlation data includes a resource consumption side statistic, a resource consumption side resource storage amount statistic, and a resource consumption side resource consumption amount statistic. The resource consumer quantity statistic value is used for counting the quantity of the resource consumers, such as the number of the medical insurance participating population corresponding to the medical insurance resource, the number of police officers corresponding to the firearm resource and the number of ordering personnel corresponding to the text resource. The resource saving amount statistic of the resource consuming party is used for counting the number of resources stored by the resource consuming party, for example, the resource saving amount statistic of the resource consuming party corresponding to the medical insurance resource includes a per-capita GDP, a per-capita dominant income, a per-capita annual wage, and the like. The resource consumption statistic of the resource consuming party is used for counting the number of resources consumed by the resource consuming party, for example, the resource consumption statistic of the resource consuming party corresponding to the medical insurance resource includes the average outpatient service cost of the medical institution, the average hospitalization cost of the medical institution, and the number of times of illness per year.

The resource service provider association data includes resource service quantity statistics. The resource service quantity statistics value is used for quantifying the resource service provided by the resource service provider, and counting the quantity of the resource service provided by the resource service provider, for example, the resource service quantity statistics value corresponding to the medical insurance resource includes the outpatient service proportion of the medical institution, the outpatient service quantity of the medical institution, the inpatient service proportion of the medical institution, the inpatient service quantity of the medical institution, and the like.

The resource service formulating party associated data comprises resource consumption service standard data and resource reservation service standard data. The resource consumption service standard data refers to a resource consumption standard established in the resource flow process and used for standardizing resource consumption services, such as medical insurance reimbursement proportion corresponding to medical insurance resources. The resource storage service standard data refers to a resource storage standard formulated in the resource flow process and used for standardizing resource storage services, such as medical insurance payment proportion, government subsidy standard and individual payment standard corresponding to medical insurance resources.

The data related to the resource storage model in the input data set comprises a resource consumption party quantity statistic value, a resource consumption party resource storage quantity statistic value and resource storage service standard data, and the resource storage result can be obtained by performing data processing on the resource consumption party quantity statistic value, the resource consumption party resource storage quantity statistic value and the resource storage service standard data through the resource storage model. The resource reservation result may reflect a global resource reservation situation.

Step S704, data processing is carried out on the data which is in the input data set and is associated with the resource service relation model through the resource service relation model, and a resource service relation result is obtained.

Specifically, the data associated with the resource service relationship model in the input data set includes resource service provider associated data, resource consumption service standard data, and resource consumer quantity statistics, and the resource service relationship result can be obtained by performing data processing on the resource service provider associated data, the resource consumption service standard data, and the resource consumer quantity statistics through the resource service relationship model. The resource service relationship result may reflect a global resource service scheduling condition.

Step S706, data associated with the resource consumption model in the input data set, the resource storage result and the resource service relationship result are processed through the resource consumption model, and a resource consumption result is obtained.

Specifically, the data associated with the resource consumption model in the input data set includes resource consumer associated data, resource service provider associated data, and resource service maker associated data. Because the resource reservation model performs data processing on part of input data to obtain a resource reservation result, and the resource service relationship model performs data processing on part of input data to obtain a resource service relationship result, the resource consumption model does not need to perform repeated processing, can directly obtain the resource reservation result and the resource service relationship result, and performs data processing on the resource reservation result, the resource service relationship result and other input data to quickly obtain the resource consumption result. The resource consumption result can reflect the global resource consumption condition.

Step S708, a service simulation result is obtained based on the resource reservation result, the resource service relationship result, and the resource consumption result.

Specifically, the computer device may obtain the service simulation result based on the resource reservation result, the resource service relationship result, and the resource consumption result, based on the numerical values among the resource reservation result, the resource service relationship result, and the resource consumption result, and the variation trend of the resource reservation result, the resource service relationship result, and the resource consumption result with respect to the historical data.

In one embodiment, obtaining the service simulation result based on the resource reservation result, the resource service relationship result, and the resource consumption result includes: when the difference between the resource storage result and the resource consumption result is smaller than a preset threshold value and the change trend of the resource service relation result meets a preset condition, determining that the service simulation result is a forward result; the trend of change is determined based on the resource service result and the historical service result; otherwise, determining that the service simulation result is a negative result.

Specifically, when the difference between the resource reservation result and the resource consumption result is smaller than the preset threshold, it indicates that the resource balance is basically balanced. And when the difference between the resource reservation result and the resource consumption result is greater than a preset threshold value, indicating that the balance and balance of resources are seriously unbalanced. When the change trend of the resource service relationship result meets the preset condition, the resource service development trend is shown to be developed towards an ideal situation. And when the variation trend of the resource service relationship result does not meet the preset condition, the resource service development trend is shown to be developed away from the ideal condition. Obviously, the basic balance of resource balance and the trend of resource service development towards the ideal situation are positive service impact results. Therefore, when the difference between the resource reserve result and the resource consumption result is smaller than the preset threshold value and the change trend of the resource service relationship result meets the preset condition, the computer device may determine that the service simulation result is a positive result, and conversely, the computer device may determine that the service simulation result is a negative result. The preset threshold value can be set according to actual needs. The trend of change of the resource service relationship result is determined based on the resource service relationship result and the historical service relationship result. It can be understood that a historical service relationship result can be obtained based on the historical input data set, a resource service relationship result can be obtained based on the current input data set, and then the resource service relationship result and the historical service relationship result are compared to obtain the change trend of the resource service relationship result.

In this embodiment, through cooperation among the resource reservation model, the resource service relationship model, and the resource consumption model, data processing can be performed on the input data set quickly to obtain a service simulation result.

In one embodiment, when the service system to be analyzed is a medical insurance service system, the plurality of subsystem models include a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relationship model, and data processing is performed on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model to obtain a service simulation result, including: performing data processing on data associated with the medical insurance resource storage model in the input data set through the medical insurance resource storage model to obtain a medical insurance resource storage result; data processing is carried out on data which are associated with the medical insurance service relation model in the input data set through the medical insurance service relation model, and a medical insurance service relation result is obtained; performing data processing on data associated with the medical insurance resource consumption model, a medical insurance resource storage result and a medical insurance service relation result in the input data set through the medical insurance resource consumption model to obtain a medical insurance resource consumption result; and obtaining a service simulation result based on the medical insurance resource storage result, the medical insurance service relation result and the medical insurance resource consumption result.

Specifically, when the service system to be analyzed is a medical insurance service system, the system dynamics model includes a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relationship model with system dynamic connection. When the system dynamics model receives the input data set and processes the data, the system dynamics model can specifically process the data associated with the medical insurance resource storage model in the input data set through the medical insurance resource storage model to obtain a medical insurance resource storage result, process the data associated with the medical insurance service relation model in the input data set through the medical insurance service relation model to obtain a medical insurance service relation result, process the data associated with the medical insurance resource consumption model in the input data set, the medical insurance resource storage result and the medical insurance service relation result through the medical insurance resource consumption model to obtain a medical insurance resource consumption result, and finally obtain a service simulation result based on the medical insurance resource storage result, the medical insurance service relation result and the medical insurance resource consumption result.

The system dynamics model corresponding to the medical insurance service system comprises a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relation model with system dynamic connection. The input data set of the medical insurance service system specifically comprises town employee associated data, town resident associated data, medical structure associated data of all levels and local government or medical insurance management department associated data.

Referring to fig. 8A, the data associated with the medical insurance resource reservation model includes town staff associated data, town resident associated data, and local government or medical insurance management department associated data, which may be the number of population of town staff insurance, the per-year wages of town staff, the individual payment proportion, and the unit payment proportion. The town resident correlation data comprise the number of population for participating in insurance of the town residents, the per-capita income of the town residents, the per-capita GDP, the individual payment standard, the government assistance standard and the like. The medical insurance income can be obtained through big data analysis and comprises the medical insurance income of urban workers and the medical insurance income of urban residents, and the medical insurance income of the urban workers is positively influenced by the number of the urban workers participating in insurance, the payment proportion and the like. The medical insurance income of the urban residents is positively influenced by the population number of the urban residents participating in insurance, the financing standard and the like, wherein the financing standard comprises medical insurance subsidy standard of the residents by the government and individual payment standard, namely government and individual financing standard. Based on the above data, a medical insurance resource reservation model as shown in fig. 8A can be established. The urban resident reference population, the urban employee reference population, the per-capitalized GDP, the urban resident per-capitalized income and the per-capitalized salary of the employees are medical insurance related data which can be obtained through computer processing and prediction. The unit payment proportion, the individual payment proportion, the government and individual financing standard and the individual financing standard are medical insurance associated data which can be obtained through user adjustment. Each arrow represents a causal relationship between data. Arrows or summaries of arrows imply a calculation formula between data. For example, the total amount of the unit payment is calculated based on the population of the employee in insurance, the payment proportion of the unit and the annual wages of the employees in the unit. The government financial subsidy income is calculated based on the urban resident reference population and the government financing standard. The medical insurance resource storage model can finally output the income of the basic medical insurance fund of the town, namely the medical insurance resource storage result.

Referring to fig. 8B and 8C, the data associated with the medical insurance service relationship model includes town employee associated data, town resident associated data, medical structure associated data of each level, and local government or medical insurance department associated data, specifically, town employee insured population, town resident insured population, number of times of illness per year, rate of visit, rate of hospitalization, medical structure visit proportion of each level, medical structure hospitalization proportion of each level, medical institution reimbursement proportion of each level, and the like. The medical insurance service effects including the outpatient quantity of the third-level medical institution, the outpatient quantity of the second-level medical institution, the outpatient quantity of the basic-level medical institution, the hospitalization quantity of the third-level medical institution, the hospitalization quantity of the second-level medical institution and the hospitalization quantity of the basic-level medical institution can be obtained through big data analysis. The total amount of the outpatients and the outpatient ratio are combined, and the outpatient amount is combined. Based on the above data, a medical insurance service relationship model as shown in fig. 8B and 8C can be established. The medical insurance service relationship model can finally output the hospitalization amount and the hospitalization amount of each stage of medical institution, namely the medical insurance service relationship result.

Referring to fig. 8D, the data associated with the medical insurance resource consumption model includes town employee associated data, town resident associated data, medical structure associated data of each level, and local government or medical insurance department associated data, specifically, town medical insurance reimbursement proportion, town resident reimbursement proportion, risk reserve proportion, and associated data affecting medical insurance service relationship and medical insurance income. The positive influence of the medical insurance consumption on the total medical cost, the reimbursement proportion and the medical insurance income can be known through big data analysis. The total medical costs are divided into total outpatient costs and total hospitalization costs, wherein the total outpatient costs are the result of the combined action of the average outpatient costs and the number of outpatients (for example, the total number of outpatients x the outpatient ratio). The total number of the patients is positively influenced by the diagnosis rate and the total number of the patients, and the total number of the patients is also influenced by the number of the insured persons and the number of the diseases per year. Based on the above data, a medical insurance service relationship model as shown in fig. 8D can be established. The medical insurance resource consumption model can finally output the expenditure of the medical insurance fund in the town, namely the medical insurance resource consumption result.

The subsystem models shown in fig. 8A, 8B, 8C, and 8D can be spliced together based on causal relationships between the associated data to obtain a complete system dynamics model.

In one embodiment, the trend of the medical service relationship result meeting the preset condition is that the treatment rate of the basic medical institution increases, so as to alleviate the current situation that the treatment pressure of the third-level medical institution is high. When the difference between the medical insurance resource storage result and the medical insurance resource consumption result is smaller than the preset threshold value, namely the medical insurance balance is achieved, and the visit rate of the primary medical institution is increased, namely the change trend of the medical service relation result meets the preset condition, the computer equipment can determine that the service simulation result is a forward result.

In one embodiment, the method further comprises: when the service simulation result is a forward result, the adjustment associated data is used as target adjustment associated data; and when the service simulation result is a negative result, updating the adjustment associated data, inputting the prediction associated data and the updated adjustment associated data into the system dynamic model until the updated service simulation result is a positive result, and taking the corresponding adjustment associated data as target adjustment associated data.

Specifically, when the service simulation result is a forward result, the computer device may take the adjustment associated data in the input data as the target adjustment associated data. And when the service simulation result is a negative result, the computer equipment updates the adjustment associated data, and inputs the prediction associated data and the updated adjustment associated data into the system dynamic model to obtain a new service simulation result. When the new service simulation result is a forward result, the computer device may take the updated adjustment associated data as the target adjustment associated data. When the new service simulation result is a negative result, the computer device may continue to update the adjustment associated data until the finally obtained service simulation result is a positive result, and use the adjustment associated data updated last time as the target adjustment associated data. The computer equipment can generate recommendation information according to the target adjustment associated data and send the recommendation information to the relevant department, and the recommendation information can provide reference information for policies and systems related to the system of the relevant department.

For example, the current adjustment associated data includes a medical insurance reimbursement proportion and a medical insurance payment proportion. When the medical insurance reimbursement proportion in the input data set is a1 and the medical insurance payment proportion is b1, if the service simulation result output by the system dynamics model is a positive result, the system indicates that the medical insurance service can be kept stable when the medical insurance reimbursement proportion is a1 and the medical insurance payment proportion is b1, and the positive development of the medical insurance service is promoted. Recommendation information can be generated according to a1 and b1 and sent to relevant departments, and the relevant departments can use a1 and b1 as reference data when setting up medical insurance policies of the next year. If the service simulation result output by the system dynamics model is a negative result, it indicates that when the medical insurance reimbursement proportion is a1 and the medical insurance payment proportion is b1, the stability of the medical insurance service cannot be maintained, and the positive development of the medical insurance service cannot be promoted. At this time, the medical insurance reimbursement proportion in the input data set can be adjusted to a2, the medical insurance payment proportion is kept to b1, if the service simulation result output again by the system dynamics model is a positive result, recommendation information can be generated according to a2 and b1 and sent to relevant departments, and the relevant departments can use a2 and b1 as reference data when making medical insurance policies of the next year.

The application also provides an application scenario, and the application scenario applies the data processing method based on the system dynamics model. Specifically, the application of the data processing method based on the system dynamics model in the application scenario is as follows:

the data processing method based on the system dynamics model can be applied to medical insurance service analysis scenes.

1. Establishing a system dynamics model corresponding to the medical insurance service system

Medical insurance is a complex system influenced by a plurality of factors such as population, economy, national policy and the like, the main body comprises governments, medical insurance departments, medical institutions, insurers and the like, and the mutual action of the factors can influence the guarantee effect. The causal relationship between the medical insurance associated data and the medical insurance associated data is determined by collecting and analyzing policy documents, documents and data related to medical insurance reimbursement and medical insurance services and determining related influence factors and factor relationships.

Through big data analysis, it can be known that health policies, population development factors and social development factors in the external environment all can influence the medical insurance service result through the data flow direction of the system dynamics model, and the medical insurance service result can be adjusted to a certain level.

Predictable and adjustable medical insurance associated data are determined from the medical insurance associated data. And establishing a system dynamic model composed of the subsystem models shown in the figures 8A, 8B, 8C and 8D based on the causal relationship among the medical insurance association data. The medical insurance related data enclosed by the rectangular frame is predictable medical insurance related data.

The system dynamics model corresponding to the medical insurance service system comprises three subsystem models, namely a medical insurance resource storage model, a medical insurance resource consumption model and a medical insurance service relation model.

2. Training data prediction model

And training a data prediction model for predicting the related medical insurance related data. And generating training samples corresponding to the data prediction models to be trained based on the historical medical insurance associated data. And carrying out supervised training on each data prediction model based on the training samples to obtain the trained data prediction model.

3. Obtaining a set of input data

Referring to fig. 9A, fig. 9A is a schematic interface diagram of the medical insurance service system. The medical insurance service system comprises a medical insurance service simulation module, a system dynamics model module and a data prediction model module. As shown in fig. 9A, a user clicks the system dynamics model module to enter a system dynamics model display interface, which includes an overall model display area, a medical insurance resource storage model display area, a medical insurance resource consumption model display area, and a medical insurance service relationship model display area. The integral model display area is used for displaying a complete system dynamics model, the medical insurance resource storage model display area is used for displaying a medical insurance resource storage model, the medical insurance resource consumption model display area is used for displaying a medical insurance resource consumption model, and the medical insurance service relation model display area is used for displaying a medical insurance service relation model. And the user can browse the corresponding display information by clicking the guide control of any display area. As shown in FIG. 9A, the user clicks on the entire model control, and the current interface shows the complete system dynamics model.

As shown in fig. 9B, the user clicks on the data prediction model module, and may enter a data prediction model display interface, where each data prediction model display area is displayed. The user can input corresponding years in the year forecasting input box, then click the forecasting control, and each data forecasting model display area can display corresponding forecasting results, so that forecasting related data can be obtained. Each data prediction model display area can also display medical insurance related data of all the years.

Referring to fig. 9C, the user clicks the medical insurance service simulation module, and may enter a medical insurance service simulation display interface, which displays the data adjustment area and the result display area. The data adjusting area is used for adjusting corresponding medical insurance related data to obtain adjusted related data. As shown in fig. 9C, the data adjustment area displays adjustment units for adjusting the relevant data, such as adjustment units corresponding to the relevant data of employees and adjustment units corresponding to the relevant data of urban residents. Some adjustment units may display relevant data of the current medical insurance policy, for example, as shown in fig. 9C, the personal payment standard of the town residents in the current medical insurance policy is 40 yuan. The user can input the relevant data of the new medical insurance policy in each adjusting unit, for example, as shown in fig. 9C, the personal payment standard of the town residents in the new medical insurance policy is adjusted to 50 yuan. In addition, regarding the medical insurance-related data which does not need to be adjusted basically, has small change and the like, the user can directly input or select default values or historical values at corresponding positions of the data adjusting area. Such as the outpatient scale of each level of medical facilities.

4. Medical insurance service simulation

Referring to FIG. 9D, predictive correlation data and regulatory correlation data may be derived based on historical correlation data, and an input data set of a system dynamics model may be derived based on the predictive correlation data and the regulatory correlation data. And inputting the input data set into a system dynamic model corresponding to the medical insurance service system, and performing data processing on data associated with the medical insurance resource storage model in the input data set through the medical insurance resource storage model to obtain a medical insurance resource storage result. And performing data processing on data associated with the medical insurance service relation model in the input data set through the medical insurance service relation model to obtain a medical insurance service relation result. And performing data processing on the data associated with the medical insurance resource consumption model, the medical insurance resource storage result and the medical insurance service relation result in the input data set through the medical insurance resource consumption model to obtain a medical insurance resource consumption result. And obtaining a service simulation result based on the medical insurance resource storage result, the medical insurance service relation result and the medical insurance resource consumption result.

Referring to fig. 9C, the medical insurance service simulation display interface displays the medical insurance reimbursement curve and the medical insurance service relationship list. The medical insurance reimbursement curve comprises a medical insurance fund income curve and a medical insurance fund consumption curve under an original medical insurance policy which are drawn based on medical insurance reimbursement results of past years and predicted medical insurance reimbursement results, and a medical insurance fund income curve and a medical insurance fund consumption curve under a new medical insurance policy. The medical insurance service relation list comprises a medical insurance service relation list under the original medical insurance policy and a medical insurance service relation list under the new medical insurance policy. As shown in FIG. 9C, the new medical insurance policy medical insurance fund income curve and the medical insurance fund consumption curve substantially overlap, indicating a balance of medical insurance reimbursements. The primary visit rate under the new medical insurance policy is increased compared with the primary visit rate under the original medical insurance policy, which indicates that the visit pressure of the third-level medical institution is relieved. Therefore, the medical insurance service simulation result corresponding to the new medical insurance policy is a forward result, and the related data of the new medical insurance policy can be stored and recommended.

In the embodiment, the medical insurance service is analyzed based on the system dynamics model, a large amount of medical insurance associated data can be integrated to quantify the medical insurance service, and an accurate data analysis result can be output, so that medical insurance decisions of relevant departments can be assisted, and the medical insurance decision quality is improved. In addition, relevant variables in the system dynamics model are optimized based on the machine learning model, so that the accuracy of an analysis result is improved, and the quality of medical insurance decisions is further improved.

It can be understood that, in addition to establishing the system dynamics model corresponding to the medical insurance resource service system, system dynamics models corresponding to service systems related to fund and insurance management, such as a public accumulation fund resource service system, an endowment insurance resource service system, a lost insurance resource service system, a work injury insurance resource service system, a birth insurance resource service system, and the like, can be established to predict the insurance management service result in the future insurance management environment. A system dynamics model corresponding to a service system related to device management, such as a firearm resource service system, a vehicle resource service system, and the like, may be further created to predict a device management service result in a future device management environment, for example, a management result generated by implementing a current device management system in the future device management environment, and a management result generated by implementing a new device management system in the future device management environment.

It should be understood that although the various steps in the flowcharts of fig. 2-4, 7 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-4 and 7 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternatively with other steps or at least some of the other steps.

In one embodiment, as shown in fig. 10, a data processing apparatus based on a system dynamics model is provided, and the apparatus may adopt a software module or a hardware module, or a combination of the two modules as a part of a computer device, and specifically includes: a data acquisition module 1002, a data input module 1004, and a data processing module 1006, wherein:

a data obtaining module 1002, configured to obtain an input data set corresponding to a service system to be analyzed; the input data set includes prediction correlation data and adjustment correlation data.

The data input module 1004 is used for inputting the input data set into a system dynamic model corresponding to the service system to be analyzed; the system dynamics model comprises a plurality of subsystem models with system dynamics connections; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of prediction correlation data and adjustment correlation data.

The data processing module 1006 is configured to perform data processing on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model, respectively, to obtain a service simulation result; the service simulation result comprises a data processing result obtained by processing data by different subsystem models.

For specific definition of the data processing device based on the system dynamics model, reference may be made to the above definition of the data processing method based on the system dynamics model, and details are not repeated here. The various modules in the data processing device based on the system dynamics model described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 11. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used for storing historical associated data, prediction associated data, adjustment associated data, target adjustment associated data, a system dynamics model, a data prediction model and other data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of data processing based on a system dynamics model.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 12. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a method of data processing based on a system dynamics model. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the configurations shown in fig. 11 and 12 are block diagrams of only some of the configurations relevant to the present disclosure, and do not constitute a limitation on the computing devices to which the present disclosure may be applied, and that a particular computing device may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is further provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.

In an embodiment, a computer-readable storage medium is provided, in which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.

In one embodiment, a computer program product or computer program is provided that includes computer instructions stored in a computer-readable storage medium. The computer instructions are read by a processor of a computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the steps in the above-mentioned method embodiments.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A method of data processing based on a system dynamics model, the method comprising:

acquiring an input data set corresponding to a service system to be analyzed; the input data set comprises prediction correlation data and adjustment correlation data;

inputting the input data set into a system dynamics model corresponding to the service system to be analyzed; the system dynamics model includes a plurality of subsystem models having system dynamics linkages; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of the prediction associated data and the adjustment associated data;

2. The method according to claim 1, wherein the predicted associated data includes associated data of at least two data types, and the obtaining of the input data set corresponding to the service system to be analyzed includes:

acquiring at least two historical associated data respectively corresponding to various data types; the historical associated data carries time information;

forming historical associated data sequences by using the historical associated data corresponding to the same data type according to the time information to obtain the historical associated data sequences corresponding to various data types;

acquiring data prediction models corresponding to various data types;

inputting each historical associated data sequence into a corresponding data prediction model to obtain prediction results corresponding to various data types;

the respective prediction results constitute the prediction correlation data.

3. The method according to claim 2, wherein the inputting each historical associated data sequence into a corresponding data prediction model to obtain a prediction result corresponding to each data type comprises:

in the current historical associated data sequence, determining the processing priority of each historical associated data based on the arrangement sequence of each historical associated data;

acquiring a reference state parameter corresponding to the current data type;

performing feature extraction on the historical associated data of the current processing priority based on the reference state parameters to obtain feature extraction results, taking the feature extraction results as the reference state parameters in the next round of feature extraction, and returning to the step of performing feature extraction on the historical associated data of the current processing priority based on the reference state parameters until the feature extraction of each historical associated data is completed to obtain target feature extraction results;

and performing fusion processing on the target feature extraction result to obtain a prediction result corresponding to the current data type.

4. The method according to claim 3, wherein the reference state parameters include a first state parameter and a second state parameter, and the performing feature extraction on the historical associated data of the current processing priority based on the reference state parameters to obtain a feature extraction result includes:

performing data splicing on the first state parameter and the historical associated data of the current processing priority to obtain spliced data;

acquiring a forgetting parameter, and forgetting the spliced data based on the forgetting parameter to obtain a forgetting factor;

acquiring an updating parameter, and updating the spliced data based on the updating parameter to obtain a target updating factor;

performing state updating on the second state parameter based on the forgetting factor and the target updating factor to obtain a second state updating parameter;

acquiring a prediction parameter, and performing prediction processing on the spliced data based on the prediction parameter to obtain a prediction factor;

fusing the prediction factor and the second state updating parameter to obtain a first state updating parameter;

and obtaining the feature extraction result based on the first state updating parameter and the second state updating parameter.

5. The method according to claim 4, wherein the forgetting parameters include a forgetting matrix and a forgetting constant, and performing forgetting processing on the spliced data based on the forgetting parameters to obtain a forgetting factor includes:

fusing the forgetting matrix and the splicing data to obtain a forgetting fusion result;

and obtaining the forgetting factor based on the forgetting constant and the forgetting fusion result.

6. The method according to claim 4, wherein the update parameters include a first update matrix, a first update constant, a second update matrix, and a second update constant, and the updating the splicing data based on the update parameters to obtain a target update factor includes:

performing fusion processing on the first updating matrix and the splicing data to obtain a first updating fusion result;

obtaining a first updating factor based on the first updating constant and the first updating fusion result;

fusing the second updating matrix and the splicing data to obtain a second updating and fusing result;

obtaining a second updating factor based on the second updating constant and the second updating fusion result;

and obtaining the target updating factor based on the first updating factor and the second updating factor.

7. The method according to claim 4, wherein the target update factor includes a first update factor and a second update factor, and performing a state update on the second state parameter based on the forgetting factor and the target update factor to obtain a second state update parameter comprises:

performing fusion processing on the forgetting factor and the second state parameter to obtain a second state forgetting result;

fusing the first updating factor and the second updating factor to obtain a fused updating factor;

and obtaining the second state updating parameter based on the second state forgetting result and the fusion updating factor.

8. The method of claim 4, wherein the prediction parameters comprise a prediction matrix and a prediction constant, and the predicting the merged data based on the prediction parameters to obtain the prediction factors comprises:

fusing the prediction matrix and the splicing data to obtain a prediction fusion result;

and obtaining the prediction factor based on the prediction constant and the prediction fusion result.

9. The method of claim 2, wherein the training process of the data prediction model is achieved by;

acquiring a training sample of a data prediction model to be trained currently; the training sample comprises a standard correlation data sequence and a corresponding standard prediction result;

inputting the standard associated data sequence into the current data prediction model to be trained to obtain an initial prediction result;

calculating to obtain a training loss value according to the initial prediction result and the standard prediction result;

and adjusting model parameters of the current data prediction model to be trained based on the training loss value until a convergence condition is met, so as to obtain the trained data prediction model.

10. The method of claim 1, wherein the obtaining the service simulation result by performing data processing on data associated with the subsystem models in the input data set through different subsystem models in the system dynamics model comprises:

performing data processing on data associated with the resource reservation model in the input data set through the resource reservation model to obtain a resource reservation result;

performing data processing on data associated with the resource service relationship model in the input data set through the resource service relationship model to obtain a resource service relationship result;

performing data processing on data associated with the resource consumption model in the input data set, the resource storage result and the resource service relationship result through the resource consumption model to obtain a resource consumption result;

and obtaining the service simulation result based on the resource reservation result, the resource service relationship result and the resource consumption result.

11. The method of claim 10, wherein obtaining the service simulation result based on the resource reservation result, the resource service relationship result, and the resource consumption result comprises:

when the difference between the resource reservation result and the resource consumption result is smaller than a preset threshold value and the change trend of the resource service relationship result meets a preset condition, determining that the service simulation result is a forward result; the trend of change is determined based on the resource service relationship result and the historical service relationship result;

otherwise, determining that the service simulation result is a negative result.

12. The method according to any one of claims 1 to 11, further comprising:

when the service simulation result is a forward result, the adjustment associated data is used as target adjustment associated data;

and when the service simulation result is a negative result, updating the adjustment associated data, inputting the prediction associated data and the updated adjustment associated data into the system dynamics model until the updated service simulation result is a positive result, and taking the corresponding adjustment associated data as target adjustment associated data.

13. A data processing apparatus based on a system dynamics model, the apparatus comprising:

the data acquisition module is used for acquiring an input data set corresponding to the service system to be analyzed; the input data set comprises prediction correlation data and adjustment correlation data;

the data input module is used for inputting the input data set into a system dynamic model corresponding to the service system to be analyzed; the system dynamics model includes a plurality of subsystem models having system dynamics linkages; the plurality of subsystem models comprise a resource storage model, a resource consumption model and a resource service relationship model, and the resource storage model, the resource consumption model and the resource service relationship model are causal relationship networks established based on causal relationships of the prediction associated data and the adjustment associated data;

14. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 12.

15. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 12.