CN111666494B - Clustering decision model generation method, clustering processing method, device, equipment and medium - Google Patents
Clustering decision model generation method, clustering processing method, device, equipment and medium Download PDFInfo
- Publication number
- CN111666494B CN111666494B CN202010403130.5A CN202010403130A CN111666494B CN 111666494 B CN111666494 B CN 111666494B CN 202010403130 A CN202010403130 A CN 202010403130A CN 111666494 B CN111666494 B CN 111666494B
- Authority
- CN
- China
- Prior art keywords
- individual
- sample data
- group
- clustering
- individual sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/60—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to nutrition control, e.g. diets
Abstract
The invention discloses a clustering decision model generation method, a clustering processing method, a clustering decision model generation device, a clustering processing device, a clustering decision model generation device and a clustering processing medium. The method comprises the following steps: acquiring short-term contribution and long-term contribution of individual features in individual sample data, inputting the short-term contribution and the long-term contribution to a preset contribution function, and outputting feature contribution; inputting the characteristic contribution degree into a preset reward function, outputting a reward value, defining a group knowledge value label of individual sample data as a behavior variable, and defining individual characteristics of the individual sample data as a state variable; inputting the behavior variable, the state variable and the reward value into a preset grouping decision model to be trained for training, obtaining a generated Q value, and inputting the reward value and the Q value into a loss function to obtain a loss value; and when the loss value is judged not to be reduced any more, marking the clustering decision model to be trained as a clustering decision model after training is finished. The invention also relates to a blockchain technique, and the individual sample data can be stored in blockchain nodes.
Description
Technical Field
The invention relates to the field of intelligent decision, in particular to a method, a device, equipment and a medium for generating and processing a grouping decision model.
Background
At present, in order to achieve the purposes of reducing cost, realizing personalized recommendation and improving recommendation effect, division of similar groups is a very important task for many companies, and the division of similar groups needs to be applied to many application scenarios, for example: diet optimization, exercise mode recommendation, commodity recommendation of similar groups, and the like. However, the current similar population division method has the following defects: 1. at present, most of decision-making considered group knowledge bases are insufficient; 2. if a plurality of group knowledge bases are considered when the group is divided, at least one part of group knowledge is always in conflict in the plurality of group knowledge bases in the grouping process, and a scheme for deciding the conflict part does not exist at present; the two points mentioned above will cause the problems of low accuracy, low comprehensiveness and unreasonable population division. Therefore, there is a need for a technical solution to solve the above-mentioned problems.
Disclosure of Invention
Therefore, it is necessary to provide a method, an apparatus, a device and a medium for generating a clustering decision model and clustering processing, which can accurately, comprehensively and reasonably determine a group to which individual sample data to be clustered belongs, in order to solve the above technical problems.
A method of generating a clustering decision model, comprising:
acquiring individual sample data to be grouped, and establishing at least two group knowledge bases by utilizing a preset grouping decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
calling a SHAP method to evaluate short-term contributions of individual features in the individual sample data through a predicted short-term ending model of a feature contribution evaluator, calling the SHAP method to evaluate long-term contributions of the individual features in the individual sample data through the predicted long-term ending model of the feature contribution evaluator, inputting the short-term contributions and the long-term contributions into a preset contribution function of the feature contribution evaluator, and outputting feature contribution of the individual features; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
acquiring a matching relation between the individual sample data and the associated group knowledge in each group knowledge base, marking a sample group value label for the individual sample data according to all the acquired matching relations, inputting the characteristic contribution degrees of all the individual characteristics of the individual sample data into a preset reward function, outputting the reward value of the individual sample data, defining the group knowledge value label of the individual sample data as a behavior variable, and defining the individual characteristics of the individual sample data as a state variable;
inputting the behavior variable, the state variable and the reward value of the individual sample data into a free-stable DQN network of a preset grouping decision model to be trained for training, acquiring a Q value generated by the free-stable DQN network according to the behavior variable and the state variable, and inputting the reward value and the Q value of the individual sample data into a loss function of the grouping decision model to be trained and then acquiring a loss value of the individual sample data;
and when the loss value is judged not to be reduced after the individual sample data is subjected to a preset training early-stop coefficient, marking the to-be-trained grouping decision model as a training-completed grouping decision model.
A clustering decision model generation apparatus, comprising:
the system comprises an establishing module, a clustering module and a clustering module, wherein the establishing module is used for acquiring individual sample data to be clustered and establishing at least two group knowledge bases by utilizing a preset clustering decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
the output module is used for calling a SHAP method to evaluate the short-term contribution of the individual features in the individual sample data through a predicted short-term ending model of a feature contribution evaluator, calling the SHAP method to evaluate the long-term contribution of the individual features in the individual sample data through the predicted long-term ending model of the feature contribution evaluator, inputting the short-term contribution and the long-term contribution to a preset contribution function of the feature contribution evaluator, and outputting the feature contribution of the individual features; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
a defining module, configured to obtain a matching relationship between the individual sample data and the group knowledge associated in each group knowledge base, mark a sample group value tag for the individual sample data according to all the obtained matching relationships, input the feature contribution degrees of all the individual features of the individual sample data to a preset reward function, output a reward value of the individual sample data, define the group knowledge value tag of the individual sample data as a behavior variable, and define the individual features of the individual sample data as a state variable;
the acquisition module is used for inputting the behavior variable, the state variable and the reward value of the individual sample data into a free-stable DQN network of a preset grouping decision model to be trained for training, acquiring a Q value generated by the free-stable DQN network according to the behavior variable and the state variable, and inputting the reward value and the Q value of the individual sample data into a loss function of the grouping decision model to be trained and then acquiring the loss value of the individual sample data;
and the marking module is used for marking the to-be-trained grouping decision model as a training-finished grouping decision model when the loss value is judged not to be reduced after the individual sample data is subjected to a preset training early-stop coefficient.
A method of cluster processing, comprising:
acquiring individual data of a group to be determined, and acquiring at least two grouping decision results which correspond to at least two group knowledge bases one to one in the individual data; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases;
if at least two clustering decision results are inconsistent, inputting the individual data into a clustering decision model to obtain a final clustering result output by the clustering decision model; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs;
and if at least two clustering decision results are consistent, determining that the clustering decision result is the final clustering result of the individual data.
A clustering apparatus, comprising:
the grouping decision result acquisition module is used for acquiring individual data of a group to be determined and acquiring at least two grouping decision results which correspond to at least two group knowledge bases one to one in the individual data; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases;
the input module is used for inputting the individual data into a clustering decision model if at least two clustering decision results are inconsistent, and then obtaining a final clustering result output by the clustering decision model; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs;
and the final clustering result determining module is used for determining that the clustering decision result is the final clustering result of the individual data if at least two clustering decision results are consistent.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the above-mentioned clustering decision model generation method or clustering processing method when executing the computer program.
A computer-readable storage medium, which stores a computer program that, when executed by a processor, implements the clustering decision model generating method or the clustering processing method described above.
According to the group decision model generation method, device, equipment and medium, after data processing is carried out on the individual sample data to be grouped, the problem that group knowledge associated with the same individual sample data in two group knowledge bases conflicts with each other is solved, namely, a group decision model is trained and established to identify the individual sample data to be grouped and then a sample group value label of the individual sample data is output, so that the group to which the individual sample data to be grouped belongs can be accurately and efficiently determined through the group decision model, in the identification process of the model, the group knowledge in the group knowledge bases can be comprehensively considered, and the group to which the individual sample data output after identification belongs also has the advantage of high rationality.
According to the grouping processing method, the grouping processing device, the grouping processing equipment and the grouping processing medium, the final grouping result of the individual data to be determined is accurately, efficiently, reasonably and comprehensively determined through the trained grouping decision model.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a schematic diagram of an application environment of a clustering decision model generation method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for generating a clustering decision model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a clustering decision model generating apparatus according to an embodiment of the present invention;
FIG. 4 is a flow chart illustrating a grouping processing method according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a grouping apparatus according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a computer device according to an embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The clustering decision model generation method provided by the invention can be applied to the application environment shown in fig. 1, wherein a client communicates with a server through a network. Among other things, the client may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster consisting of a plurality of servers.
In an embodiment, as shown in fig. 2, a clustering decision model generation method is provided, which is described by taking the server in fig. 1 as an example, and includes the following steps:
s10, acquiring individual sample data to be clustered, and establishing at least two group knowledge bases by using a preset clustering decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
understandably, the individual sample data to be grouped can refer to long-term tracking data of a large number of individuals in a certain application scene (one tracking frequency is recorded after the tracking data is collected every time), such as the tracking data of the large number of individuals in application scenes of diet optimization,/medication management mode/exercise mode recommendation/commodity recommendation of similar groups; and the individual sample data to be grouped has a plurality of individual characteristics, so a plurality of individual characteristics can be obtained by collecting the individual sample data in a preset time period, in the application scenario of diet optimization, the individual characteristics of identity information (including but not limited to sex, age and weight of an individual), the individual characteristics of health condition (including but not limited to BMI, body mass index, body fat rate and the like), the individual characteristics of diet condition, the individual characteristics of short-term outcome (whether the individual's next BMI is in a control range) and the individual characteristics of long-term outcome (whether the individual has died next time) in the sample data of the individual can be collected, or under the scene of a medication management mode, the individual characteristics of the inspection indexes, the individual characteristics of disease histories, the individual characteristics of medication histories, the individual characteristics of short-term curative effects and the individual characteristics of long-term curative effects in individual sample data can be collected; the group knowledge base contains group knowledge which mainly indicates the knowledge of people group division, and in the application scene of the dietary mode optimization, the group knowledge in the group knowledge base is the recommended group division mode of the dietary mode, for example, two groups of ' old people ' and ' children ' are divided when the old people need to eat more calcium-containing food and the children need to supplement various nutrients '. Specifically, for each group knowledge base, various rules of the aforementioned eating manner recommendation manner can be converted into a form of a decision tree by using a preset grouping decision tree, and it should be noted that group knowledge associated with the same individual sample data may conflict with each other, for example, the group a knowledge base and the group B knowledge base conflict with usage amounts of various diets associated with the same individual sample data.
Further, the obtaining of individual sample data to be grouped includes:
acquiring all the individual sample data of a preset time period, inquiring at least one preset individual characteristic item in a preset data atlas according to the individual sample data, and collecting the individual characteristics according to the preset individual characteristic item.
Understandably, the types of the individual sample data are different (the application scenes are different), and the preset time period is different, so that the application scenes optimized by the diet mode mentioned above can be the preset time period of one quarter; since each application scenario is different, the individual features in the individual sample data must also be different; specifically, a preset individual feature item corresponding to each application scene can be established through a preset data map, and when individual sample data in one application scene is determined to be collected, each individual feature related to the individual sample data can be generated; through this embodiment can confirm each individual feature of the individual sample data that will collect for the standardization of collecting and the efficiency nature of improvement collection.
S20, calling a SHAP method to evaluate the short-term contribution of the individual features in the individual sample data through a predicted short-term ending model of a feature contribution evaluator, calling the SHAP method to evaluate the long-term contribution of the individual features in the individual sample data through the predicted long-term ending model of the feature contribution evaluator, inputting the short-term contribution and the long-term contribution to a preset contribution function of the feature contribution evaluator, and outputting the feature contribution of the individual features; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
understandably, the characteristic contribution evaluator is established based on an XGBoost model and a SHAP method, and the characteristic contribution evaluator mainly comprises a short-term outcome prediction model and a long-term outcome prediction model (the two models are inconsistent in target prediction), and both the two models are the XGBoost model. The XGboost model utilizes an integration idea to serially generate a plurality of decision trees (wherein each decision tree is matched with the residual error of the last decision tree) during training, and utilizes all decision trees to obtain a prediction result after multiple rounds of decision making during use. During training, the XGBoost model for predicting short term outcome and the XGBoost model for predicting long term outcome both use data that is a large amount of individual sample data extracted in S10 (except for the last trace data, since the last trace data generally has no individual characteristics of short term outcome and long term outcome). The input features of the two XGboost models are all body features (non-ending) in extracted individual sample data, the XGboost model output for predicting short-term ending is individual features of short-term ending, and the XGboost model output for predicting long-term ending is individual features of long-term ending. After XGboost model training is complete, SHAP methods (defining features by way of SHAP values) may be invoked using two XGboost modelsImportance, showing how the XGBoost model finally reaches the predicted result of the prediction output, which is the above-mentioned short-term outcome and long-term outcome), the SHAP method is a classical method for measuring the feature importance of the machine learning model; under the application scenario of the dietary pattern optimization, the model for predicting the short-term outcome can be used for predicting the short-term contribution of whether the BMI of the individual is in the control range next time, and the model for predicting the long-term outcome can be used for predicting the long-term contribution of whether the individual dies next time, and the like; the predetermined Contribution function of the feature Contribution evaluator is constraint (i, j, k) ═ SHAP M1 (i,j)[k]-α*SHAP M2 (i,j)[k]Wherein α is a hyper-parameter, α is greater than 1, and α is greater than 1 because the long-term outcome result is generally more important than the short-term outcome result, so α is used to adjust the relative importance degree of the long-term outcome result and the short-term outcome result, and the application difference is that the short-term outcome result is generally defined as a benign outcome, so the larger the short-term outcome result is better, and the longer-term outcome is generally defined as a malignant outcome (for example, whether the next individual dies), so the smaller the long-term outcome result is, the better i means the individual sample data, j means the number of times of tracking the individual sample data, and k means a certain individual feature of the individual sample data; SHAP M1 To predict the short term contribution output by the short term outcome model (the predicted short term outcome model is labeled M1), SHAP M2 Long-term contributions output for the predicted short-term outcome model (the predicted long-term outcome model is labeled M2).
Further, the method for evaluating the short-term contribution of the individual features in the individual sample data is called by a SHAP method through a predicted short-term ending model of the feature contribution evaluator, and the method for evaluating the long-term contribution of the individual features in the individual sample data is called by a SHAP method through a predicted long-term ending model of the feature contribution evaluator, including:
acquiring at least two clustering decision results of the individual sample data corresponding to at least two group knowledge bases; one of the group knowledge bases comprises at least one group, and the grouping decision result is that the individual sample data belongs to one of the groups contained in the group knowledge base; one said group is associated with one said group knowledge;
if at least two clustering decision results are inconsistent, a SHAP method is called through a predicted short-term ending model of the characteristic contribution degree evaluator to evaluate the short-term contribution of the individual characteristics in the individual sample data, and a SHAP method is called through a predicted long-term ending model of the characteristic contribution degree evaluator to evaluate the long-term contribution of the individual characteristics in the individual sample data.
Understandably, the number of groups contained in the group knowledge base is related to the group knowledge contained in the group knowledge base, and the more group knowledge contained in one group knowledge base, the more group number contained in the group knowledge base.
In this embodiment, it is determined in advance whether the individual sample data needs to be further subjected to the training step after step S20, and at least two clustering decision results of at least two population knowledge bases corresponding to the individual sample data may be obtained and used to determine whether the clustering decision results are consistent, for example, in an application scenario of diet mode optimization, the population knowledge in the a population knowledge base recommends the vegetable diet usage amount of the population to the individual sample data of 80g-100g, and the population knowledge in the B population knowledge base recommends the vegetable diet usage amount of the population to the individual sample data of 30g-50g, which may result in mutual conflict due to inconsistency between the two clustering decision results. Therefore, training steps after step S20 are performed to solve the above-mentioned conflicting problems. Therefore, the embodiment can effectively solve the problem of the group to which the individual sample data belongs through a corresponding means.
In an embodiment, after obtaining at least two clustering decision results of the individual sample data corresponding to at least two of the group knowledge bases, the method further comprises:
and if at least two grouping decision results are consistent, deleting the individual sample data.
S30, obtaining a matching relation between the individual sample data and the associated group knowledge in each group knowledge base, marking a sample group value label for the individual sample data according to all the obtained matching relations, inputting the characteristic contribution degrees of all the individual characteristics of the individual sample data into a preset reward function, outputting the reward value of the individual sample data, defining the group knowledge value label of the individual sample data as a behavior variable, and defining the individual characteristics of the individual sample data as a state variable;
understandably, the embodiment in which step S10 is located is only to establish a group knowledge base associated with individual sample data, and step S20 filters individual sample data with consistent grouping decision results of group knowledge in each group knowledge base, but this embodiment is to determine a direct matching relationship between the individual sample data and the associated group knowledge base, that is, to determine a matching relationship between the individual sample data and the group knowledge in the group knowledge base from each individual feature in the individual sample data, for example, in an application scenario of diet mode optimization, the individual feature of diet condition in the individual sample data is vegetable diet usage amount 40g per day, while the group knowledge in the group a knowledge base recommends vegetable diet usage amount of the group to the individual sample data of 80g to 100g, and the group knowledge in the group B knowledge base recommends vegetable diet usage amount of the group to the individual sample data of 30g to 50g, therefore, the individual sample data is not matched with the group knowledge in the group knowledge base A and is matched with the group knowledge in the group knowledge base A, so that the sample group value labels (0,1) are marked for the individual sample data through all the obtained matching relations, and the sample group value labels (mathematical coordinate values) under other conditions can be obtained through similar principles (the label 1 is marked according to the group knowledge in the group knowledge base, and the label 0 is marked not according to the group knowledge in the group knowledge base), wherein the more the group knowledge bases exist, the more the coordinate dimensions of the labeled sample group value labels are; the preset reward function is that the reward (i, j) is sigma k∈K (Contribution (i, j +1, K) -Contribution (i, j, K)), wherein K refers to all individual features of the individual sample data (and K refers to one individual feature of the individual sample data), and i and j are consistent with the aforementioned Contribution function; the behavior variable is action in reinforcement learning, the state variable is state in reinforcement learning, reinforcement learning is realized by means of trial and error, and certain action is executed under the condition of stateAfter ion, the selection of action is optimized by the reward value obtained by interacting with the environment.
S40, inputting the behavior variable, the state variable and the reward value of the individual sample data into a free-stable DQN network of a preset grouping decision model to be trained for training, obtaining a Q value generated by the free-stable DQN network according to the behavior variable and the state variable, and inputting the reward value and the Q value of the individual sample data into a loss function of the grouping decision model to be trained and obtaining a loss value of the individual sample data; the loss function includes the reward value and the Q value;
understandably, the reinforcement learning method based on Q-learning is to use a Q value table to store each state and the Q value owned by each action under the state, but many dimensions in the state are continuous, and it is a difficult matter to store the Q value of each action of each state by using the table, while free-state DQN (using a four-layer DQN network, where the input layer and the state have the same dimension, the middle has two hidden layers with dimensions of 32 and 64, respectively, the output layer and the action have the same dimension, and all the above mentioned four layers are fully connected) fused neural network and Q-learning, so that the Q value is generated by using the neural network to solve the problem that the dimension of the state is continuous; the Q value is generated by the formula
And the loss function is
Wherein
Is the state variable at step j (where j denotes the jth step of an epasopode of the individual sample data in reinforcement learning, it can be understood that S20 mentioned above refers to the tracking number of the individual sample data), a j Is the behavior variable at step j, theta is the network parameter, r j The reward value of the individual sample data at step j is taken as the attenuation factor (the farther from the current step j, the more the attenuation factor is influenced by the current reward)The smaller), Q is the Q value, representing the expected value,
is the next state variable (state at step j + 1), a j’ In order to obtain the behavior of maximum Q value, theta-is consistent with theta and is also a network parameter, theta t-1 Is the parameter before the current training round (l is the specifiable hyper-parameter). It should be noted that, when the group division is performed on the sample data, the group division effect is generated only when the data is tracked next time of some individual sample data, so that it is necessary to perform the learning training by continuously playing back a plurality of individual sample data, however, the definition of the reward value is sensitive to the training impact of the conventional DQN model, and the loss function in the conventional DQN model is related to the maximum Q value, near convergence, the parameters of the conventional DQN model may still fluctuate, so this embodiment redefines the loss function in the conventional DQN model, therefore, the use of the loss function of the clustering decision model to be trained can reduce the subjectivity of the above-defined reward value, the function of adding similar regularization items into the loss function can reduce the problem of parameter fluctuation in reinforcement learning, so that the stability of the clustering decision model to be trained can be improved.
S50, when it is judged that the loss value of the individual sample data does not decrease after the individual sample data is subjected to a preset training early-stop coefficient, the to-be-trained grouping decision model is marked as a training-completed grouping decision model.
Understandably, after the round number training of the preset training early stop coefficient, the fact that the loss value in the to-be-trained grouping decision model does not fluctuate greatly and does not decrease any more can indicate that the to-be-trained grouping decision model tends to converge, and can also indicate that the to-be-trained grouping decision model is trained to be the grouping decision model.
In summary, the above-mentioned method for generating a clustering decision model is provided, and after data processing is performed on individual sample data to be clustered, a problem that group knowledge associated with the same individual sample data in two group knowledge bases conflicts with each other is solved, that is, a clustering decision model is trained and established to identify the individual sample data to be clustered and then output a sample group value tag of the individual sample data, so that a group to which the individual sample data to be clustered belongs can be accurately and efficiently determined through the clustering decision model, and in the identification process of the model, the group knowledge in the group knowledge bases can be more comprehensively considered, and the group to which the individual sample data output after identification belongs also has the advantage of high rationality.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In an embodiment, a clustering decision model generating device is provided, and the clustering decision model generating device corresponds to the clustering decision model generating method in the above embodiments one to one. As shown in fig. 3, the grouping decision model generating apparatus includes a building module 11, an output module 12, a defining module 13, an obtaining module 14, and a marking module 15. The functional modules are explained in detail as follows:
the establishing module 11 is used for acquiring individual sample data to be grouped and establishing at least two group knowledge bases by utilizing a preset grouping decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
an output module 12, configured to call a SHAP method through a predicted short-term outcome model of a feature contribution estimator to estimate a short-term contribution of an individual feature in the individual sample data, call a SHAP method through a predicted long-term outcome model of the feature contribution estimator to estimate a long-term contribution of the individual feature in the individual sample data, input the short-term contribution and the long-term contribution to a preset contribution function of the feature contribution estimator, and output a feature contribution of the individual feature; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
a defining module 13, configured to obtain a matching relationship between the individual sample data and the group knowledge associated in each group knowledge base, mark a sample group value tag for the individual sample data according to all the obtained matching relationships, input the feature contribution degrees of all the individual features of the individual sample data to a preset reward function, output a reward value of the individual sample data, define the group knowledge value tag of the individual sample data as a behavior variable, and define the individual features of the individual sample data as a state variable;
the obtaining module 14 is configured to input the behavior variable, the state variable, and the reward value of the individual sample data to a free-stable DQN network of a preset grouping decision model to be trained for training, obtain a Q value generated by the free-stable DQN network according to the behavior variable and the state variable, and input the reward value and the Q value of the individual sample data to a loss function of the grouping decision model to be trained, and then obtain a loss value of the individual sample data;
and the marking module 15 is configured to mark the to-be-trained grouping decision model as a training-completed grouping decision model when it is determined that the loss value of the individual sample data does not decrease after the individual sample data has undergone a preset training early-stop coefficient.
Further, the establishing module comprises:
and the collecting submodule is used for acquiring all the individual sample data of a preset time period, inquiring at least one preset individual characteristic item in a preset data map according to the individual sample data, and collecting the individual characteristics according to the preset individual characteristic item.
Further, the output module includes:
the acquisition submodule is used for acquiring at least two clustering decision results of the individual sample data corresponding to at least two group knowledge bases; one of the group knowledge bases comprises at least one group, and the grouping decision result is that the individual sample data belongs to one of the groups contained in the group knowledge base; one said group is associated with one said group knowledge;
and the evaluation submodule is used for calling a SHAP method to evaluate the short-term contribution of the individual features in the individual sample data through a predicted short-term ending model of the feature contribution degree evaluator if at least two clustering decision results are inconsistent, and calling the SHAP method to evaluate the long-term contribution of the individual features in the individual sample data through a predicted long-term ending model of the feature contribution degree evaluator.
Further, the output module further includes:
and the deleting submodule is used for deleting the individual sample data if at least two grouping decision results are consistent.
For specific limitations of the clustering decision model generation device, reference may be made to the above limitations of the clustering decision model generation method, which are not described herein again. The modules in the clustering decision model generating apparatus can be wholly or partially implemented by software, hardware and their combination. 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.
The evaluation data processing method provided by the invention can be applied to the application environment shown in fig. 1, wherein the client communicates with the server through the network. Among other things, the client may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster consisting of a plurality of servers.
In an embodiment, as shown in fig. 4, a clustering method is provided, which is described by taking the application of the method to the server in fig. 1 as an example, and includes the following steps:
s50, acquiring individual data of a group to be determined, and acquiring at least two grouping decision results corresponding to at least two group knowledge bases in the individual data; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases; understandably, the above-mentioned established groups include, but are not limited to, diet groups, sport groups, and similar commodity recommendation groups, among others.
S60, if at least two clustering decision results are inconsistent, inputting the individual data into a clustering decision model to obtain a final clustering result output by the clustering decision model; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs; understandably, the trained grouping decision model is used for outputting a group value label of the individual sample data after identifying the individual sample data, and determining the group to which the individual sample data belongs through the group value label, for example, if the obtained group value label is also (0,1) mentioned in the above step S30, it indicates that the group to which the individual sample data belongs is the corresponding determined group in the group B knowledge base, if the obtained group value label is (1, 0), it indicates that the group to which the individual sample data belongs is the corresponding determined group in the group a knowledge base, and the determined group can further recommend a suitable diet condition of three meals, exercise mode recommendation, group-like commodity recommendation and the like to the user through the determined group. It should be noted that the clustering decision model can be optimized in the using process of the model, and the reward values, the behavior variables and the state variables associated with all individual features in the next new individual data of the individual data are directly input into the clustering decision model so as to optimize the clustering decision model to adapt to new individual sample data;
and S70, if at least two clustering decision results are consistent, determining that the clustering decision result is the final clustering result of the individual data.
Understandably, when at least two clustering decision results are inconsistent, after the individual data are identified through the clustering decision model, the population value label of the individual data output by the model is obtained, and the final clustering result of the individual data is determined through the population value label, which is consistent with the principle of the step S50; when the two clustering decision results are consistent, the individual sample data can belong to the determined clusters contained in the clustering decision results at the same time.
In summary, the clustering processing method is provided, and the final clustering result to which the individual data belong is accurately, efficiently, reasonably and comprehensively determined by the trained clustering decision model for the group to be determined.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In an embodiment, a clustering device is provided, and the clustering device corresponds to the evaluation data processing method in the above embodiment one to one. As shown in fig. 5, the clustering processing apparatus includes a clustering decision result obtaining module 21, an input module 22, and a final clustering result determining module 23. The functional modules are explained in detail as follows:
the grouping decision result acquisition module 21 is configured to acquire individual data of a group to be determined, and acquire at least two grouping decision results corresponding to at least two group knowledge bases in the individual data one to one; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases;
an input module 22, configured to, if at least two of the clustering decision results are inconsistent, input the individual data into a clustering decision model, and then obtain a final clustering result output by the clustering decision model; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs;
and a final clustering result determining module 23, configured to determine that the clustering decision result is the final clustering result of the individual data if at least two clustering decision results are consistent.
For specific limitations of the clustering device, reference may be made to the above limitations of the evaluation model generation method, which are not described herein again. The modules in the group processing apparatus may be implemented in whole or in part by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent of 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. 6. The computer device includes a processor, a memory, a network interface, and a database 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 operating system and the computer program to run on the non-volatile storage medium. The database of the computer device is used for storing data involved in the clustering decision model generation method. 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 clustering decision model generation method, or the computer program is executed by a processor to implement a clustering processing method.
In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the clustering decision model generating method in the foregoing embodiment when executing the computer program, or implements the steps of the clustering processing method in the foregoing embodiment when executing the computer program.
In an embodiment, a computer readable storage medium is provided, on which a computer program is stored, which when executed by a processor implements the steps of the clustering decision model generation method in the above-described embodiments, or which when executed by a processor implements the steps of the clustering processing method in the above-described 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, databases, or other media used in embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (8)
1. A clustering method, comprising:
acquiring individual sample data to be grouped, and establishing at least two group knowledge bases by utilizing a preset grouping decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
calling a SHAP method to evaluate short-term contributions of individual features in the individual sample data through a predicted short-term ending model of a feature contribution evaluator, calling the SHAP method to evaluate long-term contributions of the individual features in the individual sample data through the predicted long-term ending model of the feature contribution evaluator, inputting the short-term contributions and the long-term contributions into a preset contribution function of the feature contribution evaluator, and outputting feature contribution of the individual features; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
acquiring a matching relation between the individual sample data and the associated group knowledge in each group knowledge base, marking a sample group value label for the individual sample data according to all the acquired matching relations, inputting the characteristic contribution degrees of all the individual characteristics of the individual sample data into a preset reward function, outputting the reward value of the individual sample data, defining the group knowledge value label of the individual sample data as a behavior variable, and defining the individual characteristics of the individual sample data as a state variable;
inputting the behavior variable, the state variable and the reward value of the individual sample data into a four-layer DQN network in a free-stable DQN network of a preset grouping decision model to be trained for sequential training, acquiring an output layer of the free-stable DQN network to generate a Q value according to the behavior variable and the state variable, and inputting the reward value and the Q value of the individual sample data into a loss function of the preset grouping decision model to be trained to acquire a loss value of the individual sample data; the four layers of DQN networks are respectively an input layer, two hidden layers and an output layer, and the four layers of DQN networks are fully connected;
when the loss value is judged not to be reduced after the individual sample data is subjected to a preset training early stop coefficient, marking the preset grouping decision model to be trained as a training-completed grouping decision model;
acquiring individual data of a group to be determined, and acquiring at least two grouping decision results which correspond to at least two group knowledge bases one to one in the individual data; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases; the defined population comprises a plurality of dietary populations; the individual data refers to individual characteristics corresponding to the determined population, and the individual characteristics comprise individual characteristics of health conditions, individual characteristics of diet conditions, individual characteristics of short-term fates and individual characteristics of long-term fates; the group knowledge in the group knowledge base is a group division mode recommended by a diet mode;
if at least two clustering decision results are inconsistent, inputting the individual data into the clustering decision model to obtain a final clustering result output by the clustering decision model; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs;
and if at least two clustering decision results are consistent, determining that the clustering decision result is the final clustering result of the individual data.
2. The clustering method according to claim 1, wherein the obtaining of individual sample data to be clustered comprises:
acquiring all the individual sample data of a preset time period, inquiring at least one preset individual characteristic item in a preset data atlas according to the individual sample data, and collecting the individual characteristics according to the preset individual characteristic item.
3. The clustering method according to claim 1, wherein the estimating the short-term contribution of the individual feature in the individual sample data by the SHAP method invoked by the predicted short-term outcome model of the feature contribution evaluator and the estimating the long-term contribution of the individual feature in the individual sample data by the SHAP method invoked by the predicted long-term outcome model of the feature contribution evaluator comprises:
acquiring at least two clustering decision results of the individual sample data corresponding to at least two group knowledge bases; one of the group knowledge bases comprises at least one group, and the grouping decision result is that the individual sample data belongs to one of the groups contained in the group knowledge base; one said group is associated with one said group knowledge;
if at least two clustering decision results are inconsistent, a SHAP method is called through a predicted short-term ending model of the characteristic contribution degree evaluator to evaluate the short-term contribution of the individual characteristics in the individual sample data, and a SHAP method is called through a predicted long-term ending model of the characteristic contribution degree evaluator to evaluate the long-term contribution of the individual characteristics in the individual sample data.
4. The clustering method according to claim 3, wherein after obtaining at least two clustering decision results of the individual sample data corresponding to at least two of the group knowledge bases, further comprising:
and if at least two grouping decision results are consistent, deleting the individual sample data.
5. A clustering apparatus, comprising:
the system comprises an establishing module, a clustering module and a clustering module, wherein the establishing module is used for acquiring individual sample data to be clustered and establishing at least two group knowledge bases by utilizing a preset clustering decision tree; each group knowledge base comprises group knowledge related to all the individual sample data, and the individual sample data comprises a plurality of individual features; group knowledge in at least two group knowledge bases associated with the same individual sample data conflicts with each other;
the output module is used for calling a SHAP method to evaluate the short-term contribution of the individual features in the individual sample data through a predicted short-term ending model of a feature contribution evaluator, calling the SHAP method to evaluate the long-term contribution of the individual features in the individual sample data through the predicted long-term ending model of the feature contribution evaluator, inputting the short-term contribution and the long-term contribution to a preset contribution function of the feature contribution evaluator, and outputting the feature contribution of the individual features; the characteristic contribution evaluator is established based on a SHAP method of an XGboost model;
a defining module, configured to obtain a matching relationship between the individual sample data and the group knowledge associated in each group knowledge base, mark a sample group value tag for the individual sample data according to all the obtained matching relationships, input the feature contribution degrees of all the individual features of the individual sample data to a preset reward function, output a reward value of the individual sample data, define the group knowledge value tag of the individual sample data as a behavior variable, and define the individual features of the individual sample data as a state variable;
the acquisition module is used for inputting the behavior variable, the state variable and the reward value of the individual sample data into a four-layer DQN network in a free-stable DQN network of a preset grouping decision model to be trained for sequential training, acquiring a Q value generated by the free-stable DQN network according to the behavior variable and the state variable, and inputting the reward value and the Q value of the individual sample data into a loss function of the preset grouping decision model to be trained to acquire a loss value of the individual sample data; the four layers of DQN networks are respectively an input layer, two hidden layers and an output layer, and the four layers of DQN networks are fully connected;
the marking module is used for marking the preset grouping decision model to be trained as a training-finished grouping decision model when the loss value is judged not to be reduced after the individual sample data is subjected to a preset training early-stop coefficient;
the grouping decision result acquisition module is used for acquiring individual data of a group to be determined and acquiring at least two grouping decision results which correspond to at least two group knowledge bases one to one in the individual data; one group knowledge base comprises at least one determined group, and the grouping decision result means that the individual data belong to one determined group contained in the group knowledge base; associating one of said determined populations with a population knowledge in one of said population knowledge bases; the defined population comprises a plurality of dietary populations; the individual data refers to individual characteristics corresponding to the determined population, and the individual characteristics comprise individual characteristics of health conditions, individual characteristics of diet conditions, individual characteristics of short-term fates and individual characteristics of long-term fates; the group knowledge in the group knowledge base is a group division mode recommended by a diet mode;
the input module is used for inputting the individual data into the clustering decision model to obtain a final clustering result output by the clustering decision model if at least two clustering decision results are inconsistent; the final clustering result is associated with a population value label of the individual data to which the final clustering result belongs;
and the final clustering result determining module is used for determining that the clustering decision result is the final clustering result of the individual data if at least two clustering decision results are consistent.
6. The clustering apparatus according to claim 5, wherein the output module comprises:
the acquisition submodule is used for acquiring at least two clustering decision results of the individual sample data corresponding to at least two group knowledge bases; one of the group knowledge bases comprises at least one group, and the grouping decision result is that the individual sample data belongs to one of the groups contained in the group knowledge base; one said group is associated with one said group knowledge;
and the evaluation submodule is used for calling a SHAP method to evaluate the short-term contribution of the individual features in the individual sample data through a predicted short-term ending model of the feature contribution degree evaluator if at least two clustering decision results are inconsistent, and calling the SHAP method to evaluate the long-term contribution of the individual features in the individual sample data through a predicted long-term ending model of the feature contribution degree evaluator.
7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the grouping processing method according to any one of claims 1 to 4 when executing the computer program.
8. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the clustering method according to any one of claims 1 to 4.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010403130.5A CN111666494B (en) | 2020-05-13 | 2020-05-13 | Clustering decision model generation method, clustering processing method, device, equipment and medium |
| PCT/CN2020/098829 WO2021139106A1 (en) | 2020-05-13 | 2020-06-29 | Grouping decision-making model generation method and apparatus, grouping processing method and apparatus, and device and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010403130.5A CN111666494B (en) | 2020-05-13 | 2020-05-13 | Clustering decision model generation method, clustering processing method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111666494A CN111666494A (en) | 2020-09-15 |
| CN111666494B true CN111666494B (en) | 2022-08-12 |
Family
ID=72382560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010403130.5A Active CN111666494B (en) | 2020-05-13 | 2020-05-13 | Clustering decision model generation method, clustering processing method, device, equipment and medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111666494B (en) |
| WO (1) | WO2021139106A1 (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112115322B (en) * | 2020-09-25 | 2024-05-07 | 平安科技（深圳）有限公司 | User grouping method, device, electronic equipment and storage medium |
| CN112331254A (en) * | 2020-11-24 | 2021-02-05 | 北京泽石科技有限公司 | Method and device for generating trial and error table of nonvolatile memory |
| CN112819527A (en) * | 2021-01-29 | 2021-05-18 | 百果园技术(新加坡)有限公司 | User grouping processing method and device |
| CN112836105B (en) * | 2021-02-05 | 2022-05-24 | 浙江工业大学 | Large-scale student aerobic capacity clustering method based on movement physiological characterization fusion |
| CN113763111A (en) * | 2021-02-10 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Article collocation method, device and storage medium |
| CN113076486B (en) * | 2021-04-29 | 2023-07-25 | 平安科技（深圳）有限公司 | Drug information pushing method, device, computer equipment and storage medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009109009A1 (en) * | 2008-03-06 | 2009-09-11 | Lightradar Pty. Limited | Facilitating relationships and information transactions |
| US10546099B2 (en) * | 2014-10-15 | 2020-01-28 | Brighterion, Inc. | Method of personalizing, individualizing, and automating the management of healthcare fraud-waste-abuse to unique individual healthcare providers |
| CN109255499B (en) * | 2018-10-25 | 2021-12-07 | 创新先进技术有限公司 | Complaint and complaint case processing method, device and equipment |
| CN109801705A (en) * | 2018-12-12 | 2019-05-24 | 平安科技（深圳）有限公司 | Treat recommended method, system, device and storage medium |
| CN110956224B (en) * | 2019-08-01 | 2024-03-08 | 平安科技（深圳）有限公司 | Evaluation model generation and evaluation data processing method, device, equipment and medium |
| CN110929752B (en) * | 2019-10-18 | 2023-06-20 | 平安科技（深圳）有限公司 | Grouping method based on knowledge driving and data driving and related equipment |
| CN111105266B (en) * | 2019-11-11 | 2023-10-27 | 建信金融科技有限责任公司 | Client grouping method and device based on improved decision tree |
-
2020
- 2020-05-13 CN CN202010403130.5A patent/CN111666494B/en active Active
- 2020-06-29 WO PCT/CN2020/098829 patent/WO2021139106A1/en active Application Filing
Also Published As
| Publication number | Publication date |
|---|---|
| CN111666494A (en) | 2020-09-15 |
| WO2021139106A1 (en) | 2021-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111666494B (en) | Clustering decision model generation method, clustering processing method, device, equipment and medium | |
| Zafar et al. | An effective blockchain evaluation system based on entropy-CRITIC weight method and MCDM techniques | |
| CN110060144B (en) | Method for training credit model, method, device, equipment and medium for evaluating credit | |
| CN107066616A (en) | Method, device and electronic equipment for account processing | |
| CN109598095A (en) | Method for building up, device, computer equipment and the storage medium of scorecard model | |
| CN108898476A (en) | A kind of loan customer credit-graded approach and device | |
| CN110956224A (en) | Evaluation model generation method, evaluation data processing method, evaluation model generation device, evaluation data processing equipment and medium | |
| CN112559900B (en) | Product recommendation method and device, computer equipment and storage medium | |
| JP2023523029A (en) | Image recognition model generation method, apparatus, computer equipment and storage medium | |
| Amalnick et al. | An intelligent algorithm for final product demand forecasting in pharmaceutical units | |
| US20100076799A1 (en) | System and method for using classification trees to predict rare events | |
| CN109711424A (en) | A kind of rule of conduct acquisition methods, device and equipment based on decision tree | |
| CN111494964A (en) | Virtual article recommendation method, model training method, device and storage medium | |
| CN113822499A (en) | Train spare part loss prediction method based on model fusion | |
| CN108256626A (en) | The Forecasting Methodology and device of time series | |
| CN115391561A (en) | Method and device for processing graph network data set, electronic equipment, program and medium | |
| CN112215398A (en) | Power consumer load prediction model establishing method, device, equipment and storage medium | |
| CN115409153A (en) | Attention LSTM-based animal husbandry index prediction method and prediction system | |
| CN111784066A (en) | Method, system and equipment for predicting annual operation efficiency of power distribution network | |
| CN108228787A (en) | According to the method and apparatus of multistage classification processing information | |
| CN113345564B (en) | Early prediction method and device for patient hospitalization duration based on graph neural network | |
| CN115938600A (en) | Mental health state prediction method and system based on correlation analysis | |
| CN111062511A (en) | Aquaculture disease prediction method and system based on decision tree and neural network | |
| Matjašič et al. | blockmodeling: an R package for Generalized Blockmodeling | |
| CN115187312A (en) | Customer loss prediction method and system based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40032371 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |