CN113268335B

CN113268335B - Model training and execution duration estimation method, device, equipment and storage medium

Info

Publication number: CN113268335B
Application number: CN202110702501.4A
Authority: CN
Inventors: 邹芳; 李彦良; 黄鹏
Original assignee: Ping An Life Insurance Company of China Ltd
Current assignee: Ping An Life Insurance Company of China Ltd
Priority date: 2021-06-24
Filing date: 2021-06-24
Publication date: 2023-02-07
Anticipated expiration: 2041-06-24
Also published as: CN113268335A

Abstract

The application provides a model training and execution duration pre-estimation method, a device, equipment and a storage medium, wherein the method comprises the following steps: respectively reading and analyzing the sample configuration files corresponding to the sample processes to execute the sample processes to obtain execution records corresponding to the sample processes; analyzing each sample configuration file and the corresponding execution record to extract first characteristic data of each sample configuration file and second characteristic data of each execution record, wherein the second characteristic data comprises process execution duration; generating a training sample according to the first characteristic data and the corresponding second characteristic data to obtain a training sample set; and training the execution duration estimation model to be trained by utilizing the training sample set to obtain the trained execution duration estimation model. The method and the device simplify the computation complexity of the execution duration of the RPA process, and improve the accuracy and precision. And a large reference value is provided for the aspect of a scheduling algorithm.

Description

Model training and execution duration estimation method, device, equipment and storage medium

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a model training and execution duration pre-estimation method, device, equipment and storage medium.

Background

RPA (Robotic Process Automation) is gaining wider use. In several studies on the scheduling method of the RPA tasks on different machines, the execution time required by the RPA task execution flow is a very important input variable or critical factor in the scheduling of the RPA tasks in many cases. Many scheduling algorithms are not normally used if there is no execution duration as an input, including classical scheduling algorithms such as short job priority, and many other self-developed algorithms are limited in use. In the prior art, the research on task scheduling in the RPA field is less, the scheme for estimating the execution time of the RPA task flow is also less, and the estimation accuracy and precision are not high.

Disclosure of Invention

The method aims to solve the technical problems that in the prior art, mature and effective RPA task execution time length prediction schemes are few, the execution time length prediction is inaccurate, and accurate prediction is difficult. The application provides a method, a device, equipment and a storage medium for model training and execution duration pre-estimation, and mainly aims to simplify a calculation method for RPA task flow execution duration and improve accuracy and precision.

In order to achieve the above object, the present application provides a training method of an execution duration estimation model, which includes the following steps:

respectively reading and analyzing the sample configuration files corresponding to the sample flows to execute the sample flows to obtain execution records corresponding to the sample flows;

analyzing each sample configuration file and the corresponding execution record to extract first characteristic data of each sample configuration file and second characteristic data of each execution record, wherein the second characteristic data comprises process execution duration;

generating a training sample according to the first characteristic data and the corresponding second characteristic data to obtain a training sample set;

and training the execution duration estimation model to be trained by utilizing the training sample set to obtain the trained execution duration estimation model.

In addition, in order to achieve the above object, the present application further provides a method for estimating an execution duration, where the method for estimating an execution duration includes the following steps:

acquiring and analyzing a target configuration file of a process to be estimated to extract first characteristic data of the process to be estimated;

after data processing is carried out on first characteristic data of the process to be estimated, the first characteristic data are input into a trained target execution duration estimation model to obtain the execution duration of the process to be estimated;

the trained target execution duration estimation model is obtained according to any one of the training methods of the execution duration estimation model.

In addition, in order to achieve the above object, the present application further provides a training device for an execution duration estimation model, the training device for the execution duration estimation model comprising:

the execution module is used for respectively reading and analyzing the sample configuration files corresponding to the sample flows so as to execute the sample flows and obtain execution records corresponding to the sample flows;

the first extraction module is used for analyzing each sample configuration file and the corresponding execution record so as to extract first characteristic data of each sample configuration file and second characteristic data of each execution record, wherein the second characteristic data comprises flow execution duration;

the sample generation module is used for generating a training sample according to the first characteristic data and the corresponding second characteristic data to obtain a training sample set;

and the training module is used for training the execution duration estimation model to be trained by utilizing the training sample set to obtain the trained execution duration estimation model.

In addition, to achieve the above object, the present application further provides an execution duration estimation device, where the execution duration estimation device includes:

the second extraction module is used for acquiring and analyzing a target configuration file of the process to be estimated so as to extract first characteristic data of the process to be estimated;

the estimation module is used for inputting the first characteristic data of the process to be estimated into the trained target execution duration estimation model after data processing is carried out on the first characteristic data so as to obtain the execution duration of the process to be estimated;

the trained target execution duration estimation model is obtained according to the training device of the execution duration estimation model.

To achieve the above object, the present application also provides a computer device comprising a memory, a processor and computer readable instructions stored on the memory and executable on the processor, the processor performing the steps of the method according to any one of the preceding claims when executing the computer readable instructions.

To achieve the above object, the present application also provides a computer readable storage medium having stored thereon computer readable instructions, which, when executed by a processor, cause the processor to perform the steps of the method according to any of the preceding claims.

According to the model training and execution time estimation method, the device, the equipment and the storage medium, the sample flow is executed according to the sample configuration file to obtain the training sample, and the execution time estimation model is automatically trained through machine learning based on the training sample to obtain the trained execution time estimation model. The trained execution time length estimation model can be used for estimating the execution time length of the unexecuted RPA process, so that the calculation complexity of the execution time length of the RPA process is simplified, and the accuracy and the precision are improved. And a large reference value is provided for the aspect of a scheduling algorithm. The estimated flow time can also provide estimated ending time for the flow, and better support is provided for business personnel.

Drawings

Fig. 1 is an application scenario diagram of a training method of an execution duration estimation model and an estimation method of an execution duration in an embodiment of the present application;

FIG. 2 is a schematic flow chart illustrating a training method for executing a duration estimation model according to an embodiment of the present disclosure;

FIG. 3 is a schematic flow chart illustrating a method for estimating an execution duration according to an embodiment of the present application;

FIG. 4 is a block diagram of a training apparatus for executing a duration prediction model according to an embodiment of the present disclosure;

FIG. 5 is a block diagram illustrating an apparatus for estimating an execution duration according to an embodiment of the present disclosure;

fig. 6 is a block diagram illustrating an internal structure of a computer device according to an embodiment of the present application.

The implementation, functional features and advantages of the object of the present application will be further explained with reference to the embodiments, and with reference to the accompanying drawings.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. 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 application. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.

The training method of the execution duration estimation model and the estimation method of the execution duration provided by the application can be applied to the application environment shown in fig. 1, wherein the terminal device can be but is not limited to various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices.

Of course, the training method of the execution duration estimation model and the estimation method of the execution duration provided by the application can be realized on the same terminal device, and can also be realized on different terminal devices. Different terminal devices can be connected through wired or wireless communication to carry out data transmission.

In addition, the training method of the execution duration estimation model and the estimation method of the execution duration provided by the application can be completed by an execution duration estimation system consisting of the terminal equipment and the server. The terminal device is connected with the server through a network.

Fig. 2 is a flowchart illustrating a training method for executing a duration estimation model according to an embodiment of the present disclosure. Referring to fig. 2, the training method of the execution duration estimation model is applied to the terminal device in fig. 1 as an example for explanation. The training method of the execution duration estimation model comprises the following steps S100, S300, S500 and S700.

S100: and respectively reading and analyzing the sample configuration files corresponding to the sample flows to execute the sample flows to obtain execution records corresponding to the sample flows.

Specifically, the training method for the execution duration estimation model provided by this embodiment is generally executed by the terminal device, and can be widely applied to the RPA technical field.

RPA (robot Process Automation) is a kind of automated software tool, which can use and understand the existing applications of enterprises through a user interface, automate the regular operations based on rules, and replace people to execute office processes with high regularity and repeatability in front of computers. Human-computer interaction behavior can be simulated to perform business processes. Specifically, the RPA captures and operates an application program by using a user interface on the terminal device to simulate the interaction process of a user and a computer and execute related tasks, thereby realizing automation in the workflow. Thereby taking over the hands and replacing manual operation to independently complete certain work. Has been widely applied in the fields of manufacturing industry, energy, bank finance and the like.

The sample flow or the flow to be estimated in this embodiment may be one of a plurality of flows (subtasks) to be executed by one RPA task, or may be one RPA task. At least one operation is required to be executed when each flow is to be executed. Such as downloading reports, logging in to a website, processing reports, extracting, etc. The operation to be executed by the flow is specifically defined according to an actual application scenario, and the application is not limited.

RPA tasks or processes perform differently than other programs or processes, supporting messaging mechanisms and break mechanisms. At present, after the operation flow is started, the flow is exited only when the flow execution is finished or a preset finishing condition is reached. The execution of each process needs to occupy running resources, which affects the allocation of resources, the trend of subsequent processes, whether the subsequent whole task can be successfully executed or not, and the execution efficiency of the whole service. Therefore, it is an urgent problem to estimate the execution duration of the process or task accurately.

In the RPA application, the RPA process or task is a workflow which is required to be executed when the corresponding RPA task is called and executed by the scheduling module. The execution time of the RPA process is the time consumed by the execution of one RPA process. In general, the execution time of the RPA process is used as an input variable of the scheduling module, so that the scheduling module comprehensively considers how to schedule the RPA task by combining the execution time of the RPA process and other factors. If no RPA flow execution time is used as input, scheduling algorithms in many scheduling modules cannot be used, so that the scheduling modules cannot normally schedule RPA tasks or are limited in function.

The sample profile may configure parameters and initial settings for the sample flow. The computer reads and parses the sample configuration file, and can perform various operations in the sample flow according to various setting attributes of the sample configuration file. The sample configuration file determines how the sample process should be run and which operations to perform, which is equivalent to instructing the process to run.

The sample configuration file can be written and imported into the RPA system by a developer in advance according to a specific application scenario or service requirement. For example, in some application scenarios, the process configuration file records how many reports need to be downloaded by the corresponding RPA process, which reports need to be downloaded, how large-scale data is processed, which websites need to be logged in, and the like. The configuration files represent the steps and operations that the corresponding process needs to perform.

In one embodiment, to facilitate the extraction of feature data by a machine learning model or computer, the process profile may be designed based on the RPA characteristics and the characteristics of the machine learning model. For example, the RPA flow configuration file may be in an Excel format or an xml format, although the application is not limited thereto.

The sample process is an RPA process which needs to be normally executed and can provide training samples after the execution is finished, and the sample process is used for providing sample data for the training execution duration prediction model.

Each sample flow needs to load and parse the corresponding sample configuration file before execution, and then execute.

The RPA process configuration files corresponding to each sample RPA process may be the same or different. Preferably, the RPA flow profiles corresponding to each sample RPA flow have differences.

The method can determine key parameters influencing the operation time of the RPA process according to the execution of the historical RPA process, and obtain different RPA process configuration files corresponding to a plurality of different sample RPA processes according to different key parameter configurations. Therefore, the influence of the key parameters on the execution time is highlighted by setting the key parameters, the influence and the interference of irrelevant parameters or factors on the execution time are reduced, and the aim of better training the model is fulfilled.

In one embodiment, all sample RPA flows are RPA flows of the same classification. Namely, the execution time length estimation model of the same RPA process is trained by using the training samples obtained by the RPA processes of the same classification, and the execution time length estimation model of the trained RPA process is used for estimating the RPA processes to be estimated of the same classification. The execution time length estimation model of the same RPA process is trained by using the RPA processes of the same classification, so that the training concentration degree is higher, the execution time length estimation model of the RPA process obtained after training can reflect the attribute of the RPA process of the same classification better, and the process time length of the RPA process of the same classification can be estimated more accurately.

S300: and analyzing each sample configuration file and the corresponding execution record to extract first characteristic data of each sample configuration file and second characteristic data of each execution record.

Specifically, after each RPA flow is executed, a corresponding execution record is generated. The execution log is an execution log, and details of each step in the execution process are recorded. For example, it records how many times each step was executed, whether it was successfully executed, when it was executed, the number of retries of the flow, the total execution time of the flow, the number of error steps, and the type of each step. The types of steps are, for example, downloading, extracting, processing reports, retrying, etc. Under different application scenarios, the steps that the RPA flow needs to execute may not be the same.

First characteristic data is extracted from the sample configuration file, second characteristic data is extracted from the execution record, and the second characteristic data comprises the total execution duration of the process. The first characteristic data is data that is capable of characterizing a sample flow. For example, data may include one or more dimensions of parameters performed by the sample flow, types and times of steps performed or steps critical and indispensable to the performance, critical parameters affecting the duration of the performance, and the like. The second characteristic data includes an execution time length of the flow.

S500: and generating a training sample according to the first characteristic data and the corresponding second characteristic data to obtain a training sample set.

The first characteristic data and the second characteristic data are in one-to-one correspondence, the second characteristic data are labels of the corresponding first characteristic data, and the labels are real execution duration. Each sample process corresponds to a training sample, and the training samples corresponding to all the sample processes form a training sample set.

The samples in the training set comprise the characteristics (such as the characteristics of multiple dimensions) of the samples and the target variables with definite values, so that the machine learning model can find the rule between the target variables predicted from the characteristics of the samples, and therefore the performance of predicting the values of the target variables based on the characteristics of the samples is achieved. Specifically, the execution duration estimation model can learn and acquire the rule between the execution durations of the processes according to the first feature data extracted from the configuration file, and then estimate the execution duration (second feature data) of any one process to be estimated according to the first feature data extracted from the configuration file.

S700: and training the execution duration estimation model to be trained by utilizing the training sample set to obtain the trained execution duration estimation model.

Specifically, during the training process of the machine model, the feature data needs to be processed to obtain a data format (e.g., a feature matrix) that can be recognized by the machine model. Therefore, it is necessary to generate a corresponding first feature matrix from the first feature data and a corresponding second feature matrix from the second feature data, and since the first feature data includes features of a plurality of dimensions, the first feature matrix is in the form of a feature key-value pair (key-value). The first feature matrix and the corresponding second feature matrix are features extracted for the corresponding training samples.

Inputting the characteristics of each training sample into an execution duration estimation model to be trained, training the execution duration estimation model to be trained until the obtained execution duration estimation model meets the training ending condition, obtaining various relevant model parameters of the execution duration estimation model, and generating a model configuration file according to the relevant model parameters. The training end condition may include convergence of the trained model and/or a preset number of iterations. In the training process, model iterative training can be performed by using a training sample based on a logistic regression model structure to obtain an iterated logistic regression model. Each training model has a group of relevant model parameters in the training process, and the essence of the model training is to find a better group of relevant model parameters before the training end condition is met.

The model configuration file is used for configuring the execution duration estimation model, so that the configured execution duration estimation model becomes a trained execution duration estimation model, and then execution duration estimation is carried out on the flow to be estimated.

In a specific embodiment, the training method of the execution duration prediction model may adopt an L2-customized L2-loss supported Vector Regression method, and obtain the trained execution duration prediction model by using a Support Vector Regression model. Other types of models may be used, such as using a support polynomial regression model to derive a trained execution duration prediction model.

The specific training process is as follows: extracting the characteristics of each training sample to obtain a first characteristic matrix and a corresponding second characteristic matrix corresponding to each training sample, inputting the first characteristic matrices and the corresponding second characteristic matrices of all the training samples into the execution duration estimation model to be trained, and obtaining an execution duration prediction result corresponding to each training sample after each training of the execution duration estimation model to be trained is finished; and calculating a loss function according to the difference between the execution duration prediction result and the real execution duration corresponding to the label (second characteristic data), and determining whether to stop training according to whether the value of the loss function converges to a preset value every time. And if the value of the loss function does not converge to the preset value, continuously adjusting the model parameters of the execution duration estimation model to be trained and continuously carrying out iterative training until the value of the loss function converges to the preset value. When the value of the loss function converges to a preset value, the training of the time length estimation model to be executed is completed, and the corresponding model parameter is the currently optimal model parameter. And parameterizing the to-be-executed time length estimation model according to the model parameters to obtain a trained execution time length estimation model. Of course, the model parameters corresponding to the trained execution duration estimation model can be further optimized and updated under the condition of permission of conditions and time.

According to the method, an execution record is obtained by executing a sample process through a configuration file of the sample process, first characteristic data is extracted from the configuration file, second characteristic data is extracted from the execution record, the first characteristic data and the second characteristic data are used as training samples to train an execution duration estimation model, and then the execution duration of the unexecuted process to be estimated is predicted through the trained execution duration estimation model. The execution time of the RPA process is innovatively estimated by introducing machine learning in the RPA field. As an important input for scheduling the RPA flow to which machine to execute the decision process, a great reference value is provided for the aspect of a scheduling algorithm. In classical scheduling algorithms, such as short job first, high response ratio first, shortest remaining time first, etc., job time must be used as input. In the self-developed scheduling algorithm, if the estimated duration is used as an input, the scheduling algorithm can be designed from more dimensions and indexes, for example, the algorithm meeting the service requirement can be designed from indexes such as the shortest response time and the shortest waiting time. The estimated time in milliseconds provides possibility for the next operation, such as scheduling decision process. Therefore, the scheme is a solution for deploying the pre-estimated duration model in the RPA process system, which can be practically implemented. The method and the device overcome the problems that in the estimation mode in the prior art, proper sample data cannot be fully considered and selected, or the estimation mode adopted is improper, so that the execution time length is estimated inaccurately, and the like. And a large reference value is provided for the aspect of a scheduling algorithm. The estimated flow time can also provide estimated ending time for the flow, and better support is provided for business personnel.

In one embodiment, step S200 specifically includes the following steps:

judging whether the corresponding sample flow is effectively executed according to the execution record, and taking the effectively executed sample flow as an effective sample flow;

and analyzing the sample configuration file and the execution record corresponding to each effective sample flow to extract first characteristic data and second characteristic data of the execution record of the sample configuration file corresponding to each effective sample flow.

Specifically, in the machine learning model training process, the richer and more sufficient the training samples, the better the model can be trained. The method includes that a plurality of sample flows are provided, in the execution process of each sample RPA flow, the execution time is different, the execution environment in execution may be different, in the execution process of each sample flow, some sample flows may be successfully executed due to the change of the operation environment and the different settings of the corresponding configuration files, and some sample flows may fail to execute partial steps or fail to execute the whole flow due to the bad execution environment. For a sample flow which cannot be successfully executed at all, the execution duration of the flow in the obtained execution record is necessarily inaccurate or even has no reference value. Therefore, invalid training samples are taken as bad values to be removed, the retained sample data corresponding to the effectively executed sample process can be used for more accurately training the execution duration pre-estimation model, interference of the invalid sample data on model training is reduced, the model obtained by final training is more accurate, and the trained model can output more accurate results.

Because the execution record records details of the various steps in the flow execution process. For example, data such as how many times each step was executed, whether it was successfully executed, when it was executed, the number of retries of the flow, the total execution time of the flow, and the type of each step are recorded. Therefore, whether or not each step of the flow is successfully executed can be determined by executing the record, and thus whether or not the entire flow is efficiently executed can be determined.

In an embodiment, the determining whether the corresponding sample process is effectively executed according to the execution record specifically includes the following steps:

acquiring the abnormal proportion of the execution abnormity appearing in the execution record;

if the abnormal proportion is larger than the preset threshold value, judging that the corresponding sample flow is not effectively executed;

and if the abnormal proportion is less than or equal to the preset threshold, judging that the corresponding sample flow is effectively executed.

Specifically, the reasons for the exception include downloading, extracting, handling report exceptions, and retrying. The exception proportion is the ratio of the number of steps in which execution exceptions occur to the total number of steps executed.

The preset threshold may be set according to an actual application scenario, for example, may be set to be equal to 1%, 5%, 10%, and the application is not limited thereto.

In another embodiment, the determination of whether the corresponding sample flow is effectively executed according to the execution record may further be determined according to the flow execution duration in the execution record. More specifically, an execution time average value is calculated according to execution time after a plurality of flows of the same type are successfully and successfully executed, an execution time effective range is obtained according to the execution time average value, a sample flow with the flow execution time in the effective range is judged to be effectively executed, and a sample flow with the flow execution time exceeding the effective range is judged to be not effectively executed. For example, if the execution duration validity range is 30-40s, the samples with the execution duration less than 30s or greater than 40s are determined to be not validly executed.

In one embodiment, before step S100, the method for performing training of the duration estimation model further includes the following steps:

determining parameter information influencing execution duration according to setting parameters corresponding to historical executed processes and historical execution records;

the parameter information is provided to the user for composing a sample profile.

Specifically, in practical applications, RPA flows of the same category may be configured with different parameters and executed in a large number of application processes, and thus, a large number of setting parameters and historical execution records corresponding to the historical executed flows may be generated. Parameter information influencing the execution duration can be preliminarily determined according to the historical execution record and the set parameters, the parameter information is provided for engineering personnel, and the engineering personnel can configure and write the sample configuration file by taking the parameter information as a reference, so that the parameter setting in the sample configuration file is more targeted. For example, in some sample configuration files, the irrelevant parameters are set to be the same, the key parameters are changed, and the interference and influence of the irrelevant parameters on the execution duration are reduced, so that the execution duration estimation model can more accurately acquire the rules of the key parameters and the execution duration. The embodiment can provide more purposefulness and pertinence for engineering personnel when compiling the configuration file, so that the obtained training sample is more excellent, and the model is better trained.

FIG. 3 is a schematic flow chart illustrating a method for estimating an execution duration according to an embodiment of the present disclosure; the method for estimating the execution duration is applied to the terminal device in fig. 1 as an example for explanation. The method for estimating the execution time length comprises the following steps S200 and S400.

S200: and acquiring and analyzing a target configuration file of the flow to be estimated so as to extract first characteristic data of the flow to be estimated.

Specifically, the flow to be estimated is a new unexecuted RPA flow. The first characteristic data extracted from the target configuration file of the process to be estimated is data capable of representing the characteristics of the process to be estimated. For example, one or more of parameters of the execution of the flow to be estimated, the types and times of the executed steps or the steps that are critical and indispensable in the execution, key parameters that affect the execution time length, and the like may be included.

Since the process to be estimated is not executed, in this embodiment, the execution time of the process to be estimated is not obtained after execution, but is obtained through a trained target execution time estimation model before the process to be estimated is executed.

S400: and after data processing is carried out on the first characteristic data of the flow to be estimated, inputting the first characteristic data into a trained target execution duration estimation model so as to obtain the execution duration of the flow to be estimated.

Specifically, the first feature data is data that cannot be directly recognized or read by the machine model, and the first feature data needs to be subjected to data conversion processing to obtain a first feature matrix of the first feature data, and then the first feature matrix is input into the trained target execution duration estimation model to estimate the execution duration with the estimation process.

The target execution duration estimation model is obtained by training according to any one of the training methods of the execution duration estimation model.

After the model is trained, a configuration file of the model is generated, wherein information about the weight of the model parameters and the like is recorded.

When a new unexecuted flow to be estimated arrives, firstly starting a target execution duration estimation model to load a configuration file generated when a training model is loaded, and initializing the model. And then, analyzing the configuration file of the unexecuted estimated flow and extracting a first characteristic matrix, inputting the first characteristic matrix into the loaded target execution time estimation model, and then calculating to obtain the estimated value of the execution time of the flow to be estimated.

In one specific implementation, the training method of the target execution duration estimation model comprises the following steps:

respectively reading and analyzing the target sample configuration files corresponding to the target sample flows to execute the target sample flows to obtain target execution records corresponding to the target sample flows;

analyzing each target sample configuration file and a corresponding target execution record to extract first characteristic data of each target sample configuration file and second characteristic data of each target execution record, wherein the second characteristic data comprises a process execution duration;

generating a target training sample according to the first characteristic data and the corresponding second characteristic data to obtain a target training sample set;

and training the target execution duration estimation model to be trained by using the target training sample set to obtain the trained target execution duration estimation model.

In one embodiment, the target sample process and the process to be estimated are RPA processes corresponding to the same class of RPA tasks. The method is equivalent to a classification RPA flow corresponding to an execution duration estimation model, so that the trained model is more targeted, and the accuracy of the estimated execution duration is higher.

In one embodiment, the method for estimating the execution duration further includes the following steps:

generating a new training sample according to first characteristic data and execution duration corresponding to a process to be estimated;

and iteratively training the trained target execution time length estimation model according to the new training sample so as to update the trained target execution time length estimation model.

Specifically, the relevant model parameters obtained when the model satisfies the preset end condition are a better set of parameters, but not necessarily an optimal set. After a certain amount of new training samples are collected, iterative training can be carried out on the trained target execution duration estimation model, so that the target execution duration estimation model is more and more perfect, and the estimated execution duration is more accurate.

The method combines the RPA process with a machine learning method, and carries out execution duration estimation on the process. For the RPA process, a scheme capable of deciding the process operation process according to the configuration file is designed, so that the process characteristics can be extracted from the configuration file and applied to the duration estimation. Through the lightweight architecture design, the whole training process and the estimation process are very quick, so that the iterative training of the model can be quickly carried out, and the accuracy of the estimation model is improved. The estimated time in milliseconds provides the possibility for the next operation, such as scheduling decision process. Therefore, the scheme is a practical solution for deploying the estimated time length model in the RPA process system.

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 application.

Fig. 4 is a block diagram of a training apparatus for executing a duration prediction model according to an embodiment of the present application, where the training apparatus for executing the duration prediction model includes:

the execution module 100 is configured to read and analyze sample configuration files corresponding to the sample flows respectively to execute the sample flows, so as to obtain execution records corresponding to the sample flows;

a first extraction module 300, configured to analyze each sample configuration file and a corresponding execution record to extract first feature data of each sample configuration file and second feature data of each execution record, where the second feature data includes a process execution duration;

the sample generating module 500 is configured to generate a training sample according to the first feature data and the corresponding second feature data to obtain a training sample set;

the training module 700 is configured to train the execution duration estimation model to be trained by using the training sample set, so as to obtain a trained execution duration estimation model.

In one embodiment, the first extraction module 300 specifically includes:

the screening module is used for judging whether the corresponding sample flow is effectively executed according to the execution record and taking the effectively executed sample flow as an effective sample flow;

and the sub-extraction module is used for analyzing the sample configuration file and the execution record corresponding to each effective sample flow so as to extract first characteristic data of the sample configuration file and second characteristic data of the execution record corresponding to each effective sample flow.

In one embodiment, the screening module specifically includes:

the computing unit is used for acquiring the abnormal proportion of the execution abnormity appearing in the execution record;

and the judging unit is used for judging that the corresponding sample flow is not effectively executed if the abnormal proportion is greater than the preset threshold, and judging that the corresponding sample flow is effectively executed if the abnormal proportion is less than or equal to the preset threshold.

In one embodiment, the training device for executing the duration estimation model further includes:

the analysis module is used for determining parameter information influencing execution duration according to the setting parameters corresponding to the historical executed flow and the historical execution record;

and the providing module is used for providing the parameter information for a user to write a sample configuration file.

Fig. 5 is a block diagram illustrating a structure of an apparatus for estimating an execution duration according to an embodiment of the present application, where the apparatus for estimating an execution duration includes:

the second extraction module 200 is configured to acquire and analyze a target configuration file of the process to be estimated, so as to extract first feature data of the process to be estimated;

the estimation module 400 is configured to input the first feature data of the process to be estimated into a trained target execution duration estimation model after data processing is performed on the first feature data, so as to obtain an execution duration of the process to be estimated;

the trained target execution duration estimation model is obtained according to any one of the execution duration estimation model training devices.

In one embodiment, the means for estimating the execution duration comprises:

the second sample generation module is used for generating a new training sample according to the first characteristic data and the execution duration corresponding to the flow to be estimated;

and the second training module is used for carrying out iterative training on the trained target execution time length estimation model according to the new training sample so as to update the trained target execution time length estimation model.

The meaning of "first" and "second" in the above modules/units is only to distinguish different modules/units, and is not used to define which module/unit has higher priority or other defining meanings. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to such process, method, article, or apparatus, and such that a division of modules presented in this application is merely a logical division and may be implemented in a practical application in a further manner.

For the specific limitations of the training device for the execution duration estimation model and the estimation device for the execution duration, reference may be made to the above limitations of the training method for the execution duration estimation model and the estimation method for the execution duration, which are not described herein again. All or part of the modules in the training device of the execution duration estimation model and the estimation device of the execution duration can be realized by software, hardware and the combination 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.

Fig. 6 is a block diagram illustrating an internal structure of a computer device according to an embodiment of the present application. The computer device may specifically be the terminal device in fig. 1. As shown in fig. 6, the computer apparatus includes a processor, a memory, a network interface, an input device, and a display screen, which are connected through a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory includes a storage medium and an internal memory. The storage medium may be a nonvolatile storage medium or a volatile storage medium. The storage medium stores an operating system and also stores computer readable instructions, and when the computer readable instructions are executed by the processor, the processor can realize a training method of the execution duration estimation model and an estimation method of the execution duration. The internal memory provides an environment for the operating system and the execution of computer-readable instructions in the storage medium. The internal memory may also store computer readable instructions, which when executed by the processor, may cause the processor to perform the training method for the execution duration estimation model and the estimation method for the execution duration estimation model. The network interface of the computer device is used for communicating with an external server through a network connection. 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 a shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

In addition, the training method of the execution duration estimation model and the estimation method of the execution duration can be realized by the same computer device, and can also be realized by different computer devices.

In one embodiment, a computer device is provided, which includes a memory, a processor, and computer readable instructions (e.g., a computer program) stored on the memory and executable on the processor, and when the processor executes the computer readable instructions, the steps of the method for training the execution duration estimation model in the above embodiments are implemented, for example, steps S100, S300, S500, S700 shown in fig. 2 and other extensions of the method and related steps. Alternatively, the processor executes the computer readable instructions to implement the functions of the modules/units of the training apparatus for executing the duration estimation model in the above embodiments, for example, the functions of the execution module 100, the first extraction module 300, the sample generation module 500, and the training module 700 shown in fig. 4. To avoid repetition, further description is omitted here.

In one embodiment, a computer device is provided, which includes a memory, a processor, and computer readable instructions (e.g., a computer program) stored on the memory and executable on the processor, and when the processor executes the computer readable instructions, the steps of the method for estimating an execution time duration in the above embodiments are implemented, for example, steps S200 and S400 shown in fig. 3 and other extensions of the method and related steps. Alternatively, the processor executes the computer readable instructions to implement the functions of the modules/units of the device for estimating execution duration in the above embodiments, such as the functions of the second extraction module 200 and the estimation module 400 shown in fig. 5. To avoid repetition, further description is omitted here.

The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, the processor being the control center of the computer device and the various interfaces and lines connecting the various parts of the overall computer device.

The memory may be used to store computer readable instructions and/or modules, and the processor may implement various functions of the computer apparatus by executing or executing the computer readable instructions and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, video data, etc.) created according to the use of the cellular phone, etc.

The memory may be integrated in the processor or may be provided separately from the processor.

Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices 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 readable storage medium is provided, on which computer readable instructions are stored, and when executed by a processor, the computer readable instructions implement the steps of the method for training a duration estimation model in the above embodiments, such as the steps S100, S300, S500, S700 shown in fig. 2 and other extensions of the method and related steps. Alternatively, the computer readable instructions, when executed by the processor, implement the functions of the modules/units of the training apparatus for executing the duration estimation model in the above embodiments, such as the functions of the execution module 100, the first extraction module 300, the sample generation module 500, and the training module 700 shown in fig. 4. To avoid repetition, further description is omitted here.

In one embodiment, a computer readable storage medium is provided, on which computer readable instructions are stored, which when executed by a processor implement the steps of the method for estimating an execution duration in the above embodiments, such as the steps S200 and S400 shown in fig. 3 and extensions of other extensions and related steps of the method. Alternatively, the computer readable instructions, when executed by the processor, implement the functions of the modules/units of the estimation apparatus for the execution duration in the above embodiment, for example, the functions of the second extraction module 200 and the estimation module 400 shown in fig. 5. To avoid repetition, further description is omitted here.

Those skilled in the art will appreciate that all or part of the processes of the methods of the embodiments described above can be implemented by indicating relevant hardware through computer readable instructions, which can be stored in a computer readable storage medium, and when executed, the computer readable instructions 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 may include non-volatile and/or volatile memory, among others. 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-rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct bused dynamic RAM (DRDRAM), and bused dynamic RAM (RDRAM).

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of another identical element in a process, apparatus, article, or method comprising the element.

The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments. Through the description of the foregoing embodiments, it is clear to those skilled in the art that the method of the foregoing embodiments may be implemented by software plus a necessary general hardware platform, and certainly may also be implemented by hardware, but in many cases, the former is a better implementation. Based on such understanding, the technical solutions of the present application or portions thereof contributing to the prior art may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) as described above and includes several instructions for enabling a terminal device (which may be a mobile phone, a computer, a server, or a network device) to execute the method described in the embodiments of the present application.

The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are included in the scope of the present application.

Claims

1. A method for training an execution duration prediction model, the method comprising:

training an execution duration estimation model to be trained by using the training sample set to obtain a trained execution duration estimation model;

the analyzing each sample configuration file and the corresponding execution record to extract the first feature data of each sample configuration file and the second feature data of each execution record specifically includes:

judging whether the corresponding sample flow is effectively executed according to the execution record, taking the effectively executed sample flow as an effective sample flow,

analyzing the sample configuration file and the execution record corresponding to each effective sample flow to extract first characteristic data and second characteristic data of the execution record of the sample configuration file corresponding to each effective sample flow;

the determining whether the corresponding sample process is effectively executed according to the execution record includes:

obtaining the abnormal proportion of the execution abnormity appearing in the execution record, if the abnormal proportion is larger than a preset threshold value, judging that the corresponding sample process is not executed effectively, if the abnormal proportion is smaller than or equal to the preset threshold value, judging that the corresponding sample process is executed effectively,

or the like, or, alternatively,

obtaining the flow execution time length of the corresponding sample flow according to the execution record, calculating the execution time length mean value according to the execution time length after the successful execution of the same type of flow, determining the effective range of the corresponding execution time length according to the execution time length mean value, judging that the sample flow of which the flow execution time length is in the effective range of the corresponding execution time length is effectively executed according to the type of the sample flow, and judging that the sample flow of which the flow execution time length exceeds the effective range of the corresponding execution time length is not effectively executed.

2. The method for training the execution duration prediction model according to claim 1, wherein before the respectively reading and parsing the sample configuration files corresponding to the sample processes, the method further comprises:

determining parameter information of key parameters influencing execution duration according to setting parameters corresponding to historical executed processes and historical execution records;

providing the parameter information to a user for composing the sample profile.

3. A method for estimating execution time length is characterized in that the method comprises the following steps:

after data processing is carried out on the first characteristic data of the flow to be estimated, the first characteristic data are input into a trained target execution duration estimation model to obtain the execution duration of the flow to be estimated;

the trained target execution duration estimation model is obtained according to the training method of the execution duration estimation model according to any one of claims 1-2.

4. The method for estimating the execution time period according to claim 3, further comprising:

generating a new training sample according to first characteristic data and execution duration corresponding to a flow to be estimated;

and performing iterative training on the trained target execution duration estimation model according to the new training sample so as to update the trained target execution duration estimation model.

5. A training apparatus for executing a duration estimation model, the apparatus comprising:

the first extraction module is used for analyzing each sample configuration file and the corresponding execution record so as to extract first characteristic data of each sample configuration file and second characteristic data of each execution record, wherein the second characteristic data comprises process execution duration;

the training module is used for training the execution duration estimation model to be trained by utilizing the training sample set to obtain a trained execution duration estimation model;

wherein the first extraction module comprises:

the sub-extraction module is used for analyzing the sample configuration file and the execution record corresponding to each effective sample process so as to extract first characteristic data and second characteristic data of the execution record of the sample configuration file corresponding to each effective sample process;

the screening module is specifically configured to obtain an abnormal proportion of execution abnormality occurring in the execution record, determine that the corresponding sample flow is not executed effectively if the abnormal proportion is greater than a preset threshold, determine that the corresponding sample flow is executed effectively if the abnormal proportion is less than or equal to the preset threshold,

or the like, or, alternatively,

the screening module is specifically configured to obtain a flow execution duration corresponding to the sample flow according to the execution record, calculate an execution duration mean value according to the execution duration after the successful execution of the same type of flow, determine a corresponding execution duration effective range according to the execution duration mean value, determine, according to the type of the sample flow, that the sample flow of which the flow execution duration is within the corresponding execution duration effective range is effectively executed, and determine that the sample flow of which the flow execution duration exceeds the corresponding execution duration effective range is not effectively executed.

6. An apparatus for estimating an execution time period, the apparatus comprising:

the estimation module is used for inputting the first characteristic data of the flow to be estimated into a trained target execution duration estimation model after data processing is carried out on the first characteristic data so as to obtain the execution duration of the flow to be estimated;

the trained target execution duration estimation model is acquired by the training device of the execution duration estimation model according to claim 5.

7. A computer device comprising a memory, a processor, and computer readable instructions stored on the memory and executable on the processor, wherein the processor when executing the computer readable instructions performs the steps of the method of any one of claims 1-4.

8. A computer readable storage medium having computer readable instructions stored thereon, which, when executed by a processor, cause the processor to perform the steps of the method of any one of claims 1-4.