Operation information determination method and device based on causal inference and model training method

Through a causal inference method, the fluctuation prediction model is used to analyze the operation data of customer objects in the logistics industry, and the effective intervention actions are determined, which solves the problems of low efficiency and insufficient accuracy in the existing technology, and achieves more efficient intervention actions decisions .

CN119067220APending Publication Date: 2024-12-03BEIJING JINGDONG YUANSHENG TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411154398.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-21
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the logistics industry, the effective intervention actions used by the prior art to determine customer objects mainly rely on manual analysis, which is inefficient and insufficient accuracy.

Method used

Using an operation information determination method based on causal inference, the operation data and candidate operation information of the target object are obtained and input into a pre-trained fluctuation prediction model to predict the key parameter fluctuation amount corresponding to the candidate operation information, thereby determining the target operation information.

Benefits of technology

The relatively objective evaluation of the target object is achieved, the prediction accuracy is improved, the workload of people is reduced, and the processing efficiency is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067220A_ABST
    Figure CN119067220A_ABST
Patent Text Reader

Abstract

The invention provides an operation information determination method and device based on causal inference, a fluctuation prediction model training method and device, electronic equipment, a computer readable storage medium and a computer program product, and relates to the technical field of computers. According to the specific implementation scheme, operation data and at least one piece of candidate operation information of a target object are determined, wherein the candidate operation information is used for representing candidate intervention operation which can be implemented on the target object; inputting the operation data and the at least one piece of candidate operation information into a pre-trained fluctuation prediction model to obtain a key parameter fluctuation quantity of the target object corresponding to each piece of candidate operation information, the fluctuation prediction model being obtained by training based on a causal inference model; and determining target operation information from the at least one piece of candidate operation information based on the key parameter fluctuation quantity. The target object is evaluated relatively objectively, the prediction accuracy is high, the manual work amount is reduced, and the processing efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method, device and model training method for determining operation information based on causal inference. Background Art

[0002] In the logistics industry, the behaviors of customer objects settled in the merchant platform during the contract period are usually diverse. In order to achieve better management, the merchant platform will implement some intervention actions on the customer objects. Therefore, how to determine the effective intervention actions for customer objects is a technical problem that needs to be solved urgently.

[0003] Existing methods for determining effective intervention actions for customer objects are usually determined through manual analysis, that is, by manually analyzing the operational data of the customer objects to determine the effective intervention actions. This method usually has problems such as low efficiency and the determined intervention actions are not accurate enough. Summary of the invention

[0004] The present disclosure provides a method and device for determining operation information based on causal inference, a method and device for training a fluctuation prediction model, an electronic device, a computer-readable storage medium, and a computer program product.

[0005] According to one aspect of the present disclosure, there is provided a method for determining operation information based on causal inference, comprising:

[0006] Determining operation data of a target object and at least one candidate operation information, wherein the candidate operation information is used to characterize a candidate intervention operation that can be performed on the target object;

[0007] Inputting the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain a fluctuation amount of a key parameter of the target object corresponding to each candidate operation information, wherein the fluctuation prediction model is trained based on a causal inference model;

[0008] Based on the key parameter fluctuation amount, target operation information is determined from the at least one candidate operation information.

[0009] According to another aspect of the present disclosure, a fluctuation prediction model training method is provided, comprising:

[0010] Acquire historical operation data and historical operation information of the target object and historical fluctuations of key parameters after the target object is subjected to an intervention operation represented by the historical operation information;

[0011] The historical operating data and historical operation information are used as input data, and the historical fluctuation amount is used as expected output data, and the initial causal inference model is trained to obtain a fluctuation prediction model.

[0012] According to another aspect of the present disclosure, there is provided an operation information determination device based on causal inference, comprising:

[0013] an information collection module configured to determine operation data of a target object and at least one candidate operation information, wherein the candidate operation information is used to characterize a candidate intervention operation that can be implemented on the target object;

[0014] A prediction module is configured to input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain a fluctuation amount of a key parameter of the target object corresponding to each candidate operation information, wherein the fluctuation prediction model is trained based on a causal inference model;

[0015] The determination module is configured to determine target operation information from the at least one candidate operation information based on the key parameter fluctuation amount.

[0016] According to another aspect of the present disclosure, a fluctuation prediction model training device is provided, comprising:

[0017] A feature acquisition module configured to acquire historical operation data and historical operation information of a target object and historical fluctuations of key parameters of the target object after an intervention operation represented by the historical operation information is performed on the target object;

[0018] The training module is configured to use the historical operating data and historical operation information as input data and the historical fluctuation amount as expected output data to train the initial training model to obtain a fluctuation prediction model.

[0019] According to another aspect of the present disclosure, there is provided an electronic device, including:

[0020] at least one processor; and

[0021] a memory communicatively connected to the at least one processor; wherein,

[0022] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any embodiment of the above method.

[0023] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute any one of the above method embodiments.

[0024] According to yet another aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the above-mentioned method embodiments.

[0025] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0027] Figure 1 is an exemplary system architecture diagram to which the embodiments of the present disclosure may be applied;

[0028] Figure 2 is a schematic diagram of a first embodiment of a method for determining operation information based on causal inference according to the present disclosure;

[0029] Figure 3 is a schematic diagram of an application scenario of an embodiment of a method for determining operation information based on causal inference according to the present disclosure;

[0030] Figure 4 is a schematic diagram of a second embodiment of a method for determining operation information based on causal inference according to the present disclosure;

[0031] Figure 5 is a schematic diagram of a first embodiment of a fluctuation prediction model training method according to the present disclosure;

[0032] Figure 6 is a schematic diagram of a second embodiment of a fluctuation prediction model training method according to the present disclosure;

[0033] Figure 7 is a structural schematic diagram of an embodiment of an operation information determination device based on causal inference according to the present disclosure;

[0034] Figure 8 is a structural schematic diagram of an embodiment of a fluctuation prediction model training device according to the present disclosure;

[0035] Fig. 9 It is a block diagram of an electronic device used to implement the method for determining operation information based on causal inference or the method for training a fluctuation prediction model according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0036] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0037] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the causal inference-based operation information determination method, apparatus, model training method, and model training apparatus of the present disclosure can be applied.

[0038] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102 and server 103. Network 102 is used to provide a medium for a communication link between terminal 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0039] The terminal device 101 interacts with the server 103 via the network 102 to receive or send messages, etc. Various client applications may be installed on the terminal device 101 to interact with the server 103 .

[0040] It should be noted that the server 103 can be hardware or software. When the server 103 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server 103 is software, it can be implemented as multiple software or software modules (for example, for providing distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0041] It should be noted that the operation information determination method, device, model training method and model training device based on causal inference provided in the embodiments of the present disclosure can be executed by the server 103 or by the terminal device 101, and the present disclosure does not impose any restrictions on this.

[0042] Figure 2 200 is a schematic diagram of a first embodiment of a method for determining operation information based on causal inference according to the present disclosure. Figure 2 The method for determining operation information based on causal inference provided by the embodiment of the present disclosure includes:

[0043] Step S201: Determine the operation data of the target object and at least one candidate operation information, where the candidate operation information is used to represent a candidate intervention operation that can be implemented on the target object.

[0044] Step S202: input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain the fluctuation amount of the key parameters of the target object corresponding to each candidate operation information, and the fluctuation prediction model is determined based on the causal inference model.

[0045] In the embodiments of the present disclosure, the target object refers to a seller that has settled in the platform, such as a third-party seller, and the operation data refers to data that can cause fluctuations in the key parameters of the target object. In some embodiments, the operation data may include at least one of the target object's order data, waybill data, evaluation feedback data, frequency of attention data, time feature data, and portrait data.

[0046] Among them, order data may include the amount of a single order and the total order volume; waybill data includes logistics-related information, such as product identification code, category and route; evaluation feedback data may include at least one unstructured text information such as visit information to the target object, complaint information received by the target object and operational quality monitoring information of the target object. In practical applications, these unstructured text information can be segmented into different levels of features by performing Chinese word segmentation and structure tree type classification. Evaluation feedback data can increase the data diversity input into the volatility prediction model, thereby improving the prediction accuracy; the frequency of attention data includes the number or frequency of at least one operation performed by buyers on the target object, including browsing, clicking, consulting and adding to cart; time feature data represents the time when the order occurs. For example, the time feature can be the holiday time, promotion time or the corresponding year, month and week time when the order occurs; portrait data may include the location of the target object, tertiary industry, main categories and contract products.

[0047] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0048] The candidate operation information represents the candidate intervention operations that can be implemented on the target object. For example, the intervention action may include at least one of an incentive action, a renewal execution action, an activation action, and a discount contract signing action implemented on the target object. Each intervention action is configured with corresponding operation information. For example, the operation information corresponding to the incentive action can be configured as the number 1, the operation information corresponding to the renewal execution action can be configured as the number 2, the operation information corresponding to the activation action can be configured as 3, and the operation information corresponding to the discount contract signing action can be configured as 4.

[0049] In the embodiment of the present disclosure, the operation data of the target object acquired by the execution subject may be operation data within a preset time period. For example, the preset time period may be one day or one week.

[0050] After obtaining the operating data of the target object and determining at least one candidate operation information, the operating data and the candidate operation information are input into a pre-trained fluctuation prediction model, which can predict the fluctuation amount of the key parameters of the target object after the intervention action corresponding to the candidate operation information is executed. The key parameters refer to parameters that can characterize the value of the target object. In some embodiments, the target parameters can be either order volume or revenue.

[0051] In the disclosed embodiment, the fluctuation prediction model is obtained based on the training of the causal inference model. The causal inference model is a statistical model that infers and estimates causal relationships in observed data and has good prediction capabilities. In some implementations of this embodiment, the causal inference model adopts a double machine learning (DML) model. The DML model can further infer the relationship between the observed data and the output data based on the influence of the control data (treatment variables) on the observed data and the output data, and can establish the model based on more complex input data. The fluctuation prediction model trained based on the DML model in this application can more accurately characterize the corresponding relationship between the target object's operating data, the intervention actions received, and the fluctuation amount of key parameters.

[0052] In some embodiments, operational data can be divided into variable data and immutable data, wherein variable data can be data that changes over time, for example, it can include at least one of the order data, waybill data, evaluation feedback data, attention frequency data, and time feature data of the target object; immutable data can be data that is basically independent of time changes and can remain stable, for example, it can include the portrait data of the target object. By dividing operational data into variable data and immutable data, when the operational data of the target object is determined again in the future, only the variable data part can be updated while the immutable data part remains unchanged, thereby reducing the amount of calculation of the execution subject and improving the efficiency of data acquisition.

[0053] In some embodiments, step S202 specifically includes: for any candidate operation information, splicing the variable data, the immutable data and the candidate operation information to obtain target feature data; inputting the target feature data into a pre-trained fluctuation prediction model to obtain the key parameter fluctuation amount of the target object corresponding to the candidate operation information.

[0054] In practical applications, after determining the variable data, immutable data, and candidate operation information, the pipeline mechanism can be used to serialize all the data and finally aggregate them into a target feature data. The target data is input into the pre-trained fluctuation prediction model to obtain the fluctuation of the key parameters of the target object corresponding to the candidate operation information. By using the pipeline mechanism for feature splicing, repetitive data processing, normalization, and classification operations can be avoided in the process of generating feature data, thereby improving the efficiency of feature generation and avoiding repeated design code.

[0055] Step S203: Determine target operation information from at least one candidate operation information based on the fluctuation amount of the key parameter.

[0056] After determining the fluctuation amount of the key parameter, the execution subject determines the target operation information from at least one candidate operation information according to the fluctuation amount of the key parameter. It should be noted that in the embodiment of the present disclosure, when there are multiple candidate operation information, the operation data is combined with each candidate operation information as the input of the fluctuation prediction model, so as to obtain the fluctuation amount of the key parameter for each candidate operation information, and then determine the target operation information according to the fluctuation amount of the key parameter.

[0057] In some embodiments, the target operation information can be determined by judging whether the key parameter fluctuation amount meets the preset conditions. For example, when the key parameter fluctuation amount is a positive number, it can be judged whether the key parameter fluctuation amount is greater than a first preset threshold. When it is determined that the key parameter fluctuation amount is greater than the first preset threshold, it can be considered that the candidate operation information corresponding to the key parameter fluctuation amount can promote the value of the target object, for example, an increase in the order volume, thereby determining the candidate operation information as the target operation information, wherein the first preset threshold is configured as the minimum threshold representing the increase in the value of the target object; or, when the key parameter fluctuation amount is a negative number, it can be judged whether the key parameter fluctuation amount is not lower than the second preset threshold. When it is judged that the key parameter fluctuation amount is not lower than the second preset threshold, it can be considered that the candidate operation information corresponding to the key parameter fluctuation amount will not cause the value of the target object to decrease, thereby determining the candidate operation information as the target operation information, and the second preset threshold is configured as the maximum upper limit representing the decrease in the value of the target object. Taking the key parameter as order volume as an example, assuming that the first preset threshold is 80 orders, when the fluctuation of the order volume is determined to be 100 orders through the fluctuation prediction model, since 100 is greater than 80, it can be determined that the corresponding candidate operation information at this time can increase the value of the target object, and the candidate operation information is determined as the target operation information; for another example, assuming that the second preset threshold is -10 orders, when the fluctuation of the order volume is determined to be -5 orders through the fluctuation prediction model, since -5 is greater than -10, it can be considered that the corresponding candidate operation information at this time will not cause the value of the target object to decrease, and the candidate operation information can be determined as the target operation information.

[0058] In the disclosed embodiment, the target operation information may be one or more items. The target operation information indicates that the value of the target object will be enhanced after the user takes the intervention action corresponding to the target operation information on the target object. In actual applications, after determining the target operation information, the executing entity may display the intervention action indicated by the target operation information and provide relevant prompt information to facilitate the user to make a decision on the target object based on the prompt information.

[0059] The method for determining operation information based on causal inference provided by the embodiments of the present disclosure predicts the fluctuation amount of the key parameters of the target object through a pre-trained fluctuation prediction model, and then determines the target operation information according to the predicted fluctuation amount. Compared with the manual determination method in the prior art, it achieves a relatively objective evaluation of the target object and has a higher prediction accuracy, reduces the amount of manual work, and improves processing efficiency.

[0060] Figure 3 is a schematic diagram of an application scenario of the method for determining operation information based on causal inference according to this embodiment, such as Figure 3 As shown, in Figure 3In the application scenario, the execution subject 31 obtains the operation data 301 and at least one candidate operation information 302 of the target object online, inputs the operation data 301 and at least one candidate operation information 302 into the fluctuation prediction model 32 pre-trained offline, obtains the key parameter fluctuation amount 303 of the target object, and then determines the target operation information 304 according to the key parameter fluctuation amount 303.

[0061] Figure 4 400 is a schematic diagram of a second embodiment of a method for determining operation information based on causal inference according to the present disclosure. Figure 4 The method for determining operation information based on causal inference provided by the embodiment of the present disclosure includes:

[0062] Step S401: Determine the operation data of the target object and at least one candidate operation information, where the candidate operation information is used to represent a candidate intervention operation that can be implemented on the target object.

[0063] Step S402: input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain the fluctuation amount of the key parameter of the target object corresponding to each candidate operation information, and the fluctuation prediction model is determined based on the causal inference model.

[0064] The above steps S401-S402 are respectively Figure 2 Step S201 and step S202 of the illustrated embodiment correspond to each other, and therefore are not described again here. For details, please refer to the description of step S201 and step S202.

[0065] Step 403: Acquire multiple operation data of the target object in multiple time periods, and execute the above step S402 multiple times based on the multiple operation data.

[0066] In an embodiment of the present disclosure, the execution entity may obtain multiple operation data of the target object within multiple time periods, and execute the above step S402 for each operation data to obtain the key parameter fluctuation amount corresponding to each candidate operation information of the target object within the multiple time periods. For example, the execution entity may obtain the operation data of the target object daily, and based on the operation data obtained daily, input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to predict the key parameter fluctuation amount of the target object corresponding to each candidate operation information, wherein the multiple time periods refer to multiple days. In some other embodiments, the multiple time periods may also be multiple weeks, and the corresponding operation data of each time period refers to the weekly operation data of the target object. In actual applications, in order to achieve finer-grained prediction, the operation data with the smallest time granularity is usually selected for fluctuation prediction.

[0067] After obtaining a plurality of operation data of a plurality of time periods, the plurality of operation data may be stored in a preset database, so that target operation information can be determined later by querying the data in the database.

[0068] Step 404: Determine the total amount of key parameter fluctuations corresponding to each candidate operation information based on the key parameter fluctuations corresponding to each candidate operation information in multiple time periods.

[0069] After obtaining the key parameter fluctuation amount corresponding to each candidate operation information of the target object in multiple time periods, the total key parameter fluctuation amount corresponding to each candidate operation information is determined, wherein the total key parameter fluctuation amount is determined based on the key parameter fluctuation amount. In some embodiments, the total key parameter fluctuation amount is determined by summing multiple key parameter fluctuation amount.

[0070] In actual applications, the execution subject may provide a query interface or query interface for users to input query requirements. The query requirements may include a time length. After obtaining the query requirements, the execution subject extracts the fluctuation amounts of each key parameter within the time length from the preset database according to the time length included in the query requirements, and sums up the fluctuation amounts of each key parameter to obtain the total fluctuation amount of the key parameter. For example, if the time length included in the query requirements is one month, the execution subject queries the fluctuation amounts of each key parameter within one month that are pre-predicted and stored from the preset database, and sums them up to obtain the total fluctuation amount of the key parameter. It should be noted that the total fluctuation amount of the key parameter corresponds to the candidate operation information one by one, that is, each candidate operation information corresponds to a total fluctuation amount of the key parameter. In some embodiments, the query requirements may also include preset candidate operation information. At this time, the user can specifically query the total fluctuation amount of the key parameter corresponding to one or some candidate operation information, thereby realizing the screening of candidate operation information within a precise range.

[0071] Step S405: In response to the total amount of fluctuation of the key parameter satisfying a preset condition, the corresponding candidate operation information is determined as the target operation information.

[0072] After obtaining the total amount of key parameter fluctuations, the total amount of key parameter fluctuations is judged, and the target operation information is determined by judging whether the total amount of key parameter fluctuations meets the preset conditions. The setting of the preset conditions is similar to the setting in step S203 in the previous embodiment, that is, the preset condition can be that the total amount of key parameter fluctuations is greater than the first preset threshold, or it can be that the total amount of key parameter fluctuations is not less than the second preset threshold. The specific settings of the first preset threshold and the second preset threshold can be referred to the description in the above step S203, which will not be repeated here.

[0073] In response to the total amount of key parameter fluctuations satisfying the preset conditions, the corresponding candidate operation information is determined as the target operation information; in response to the total amount of key parameter fluctuations not satisfying the preset conditions, it means that the target operation information cannot be screened out. At this time, the executing entity can display the total amount of key parameter fluctuations corresponding to each candidate operation information, and give a prompt message that the target operation information cannot be determined.

[0074] Compared with the previous embodiment, the method for determining operation information based on causal inference provided by the embodiment of the present disclosure obtains operation data of multiple time periods and obtains multiple key parameter fluctuations based on the fluctuation prediction model. It can realize the prediction of key parameter fluctuations in variable time windows of different time lengths, can meet the different needs of users, and can further improve the accuracy of user decision-making.

[0075] Figure 5 FIG. 5 is a schematic diagram 500 of a first embodiment of a fluctuation prediction model training method according to the present disclosure. Figure 5 As shown, the method includes:

[0076] Step S501: Acquire historical operation data and historical operation information of a target object and historical fluctuations of key parameters after an intervention operation represented by the historical operation information is performed on the target object.

[0077] In the embodiment of the present disclosure, the historical operating data, historical operation information of the target object, and the historical fluctuations of key parameters after the intervention operation represented by the historical operation information is performed on the target object have the same meaning as the operating data, operation information and key parameter fluctuations involved in steps S201 and S202 in the aforementioned method 200, and will not be repeated here.

[0078] The acquisition of the above data can be achieved in a variety of ways. For example, the above historical data of the target object can be retrieved from the database, and the historical operating data, historical operation information and historical fluctuations of key parameters can be determined through manual sorting. Of course, in some embodiments, the historical operating data, historical operation information and historical fluctuations of key parameters can also be determined through computer execution.

[0079] As an optional implementation method, obtaining the historical operation data, historical operation information of the target object and the historical fluctuation amount of the key parameters after the intervention operation represented by the historical operation information is performed on the target object includes: determining a sampling timestamp; obtaining the operation data within a preset time period before the sampling timestamp as the historical operation data, determining the operation information corresponding to the intervention operation occurring at the sampling timestamp as the historical operation information, and determining the fluctuation amount of the key parameters of the target object within a unit time after the sampling timestamp as the historical fluctuation amount.

[0080] In this embodiment, when obtaining training data for training the fluctuation prediction model, the sampling timestamp is first determined. The sampling timestamp can be the historical moment when the intervention operation occurs. Specifically, the sampling timestamp can be determined by querying the data log of the target object. After the sampling data stamp is determined, the operation data within the preset time period before the sampling timestamp is used as the historical operation data. The preset time period can be a preset fixed value, such as one month, or the time period between the moment when the last intervention operation occurred and the sampling timestamp can also be determined as the preset time period; the fluctuation amount of the key parameters within the unit time after the sampling timestamp is determined as the historical fluctuation amount, wherein the unit time can be one day, one week or one month, and the historical fluctuation amount can be determined by calculating the change amount of the key parameters between the sampling timestamp and the prediction time. The prediction time is related to the unit time. For example, if the unit time is one day, the prediction time is the sampling timestamp plus the time corresponding to one day.

[0081] In actual applications, the execution entity may provide a data acquisition interface or interface, through which the user inputs data acquisition requirements, and the data acquisition requirements include a sampling timestamp and a unit time. After receiving the data acquisition requirements, the execution entity acquires the historical operation data, historical operation information, and historical fluctuations of key parameters of the target object after the intervention operation represented by the historical operation information is performed on the target object based on the sampling timestamp included in the data acquisition requirements. In some embodiments, the data acquisition requirements may also include the types of operation data, that is, the data acquisition requirements may also include one or more of the order data, waybill data, evaluation feedback data, attention frequency data, time feature data, and portrait data included in the operation data. The user may add the types of operation data that are expected to be acquired for training the fluctuation prediction model in the data acquisition requirements, so that the trained fluctuation prediction model can better meet the user's needs and improve the user experience.

[0082] Step S502: Using historical operation data and historical operation information as input data and historical fluctuations as expected output data, the initial causal inference model is trained to obtain a fluctuation prediction model.

[0083] In the disclosed embodiment, the initial causal inference model is a machine learning model including a three-layer structure, which are: a linear regression model, an XGB ensemble tree model (eXtreme Gradient Boosting, referred to as XGB) and a DNN model (Deep Neural Network, referred to as DNN), wherein the linear regression model is used to fit the control variable (Treatment variable), the XGB ensemble tree model is used for regression output, and the DNN model is used for embedding vector regression. The combination of the first two models is a typical paradigm of a general causal inference model, and its purpose is to obtain the average causal effect of each sample. The last layer of the model is based on DNN, and processes time series signals of arbitrary length through causal convolution, and outputs the same prediction length. In the DNN model, considering that the historical operation data used in the disclosed embodiment spans a relatively long period of time, a dilated convolution structure is adopted to increase the receptive field of each convolution layer, so as to capture the time dependency relationship of a long time ago.

[0084] The three-layer structure contained in the initial causal inference model forms an end-to-end integrated learning model, which is divided into input layer, hidden layer and output layer. During the model training process, the scalar value of the last hidden layer (the layer before the output layer) is used as the regression result. The regression result is the predicted data, which is used to calculate the loss, and then the model parameters are tuned according to the loss.

[0085] In the disclosed embodiment, the loss calculation adopts mean square error loss calculation. In order to maintain the balance of prediction results of different magnitudes, the original label data can be logarithmized so that the model can optimize parameters using the optimization method based on gradient descent; and the logarithmic processing of the label data can shrink the data dimension, thereby reducing the amount of calculation and improving the operation speed. It should be noted that if the original label data is logarithmized when training the model, when calling the trained fluctuation prediction model for fluctuation prediction, the corresponding predicted result data needs to be exponentially calculated to restore it to the actual predicted value.

[0086] The embodiment of the present disclosure can tune the model from two aspects: parameter tuning and feature tuning. Among them, parameter tuning is mainly considered from aspects such as model parameter initialization, maximum number of iterations, selection of activation function, selection of optimizer, early stopping condition, setting of learning rate, etc., and the present disclosure does not impose any restrictions on this. In addition, the embodiment of the present disclosure adopts a grid search-based method to obtain a model with optimal effect; feature tuning is to reduce the feature dimension according to the principle of Occam's razor after further analyzing the importance of features and evaluating the correlation between features, so as to avoid overfitting.

[0087] In the embodiment of the present disclosure, the input data can be randomly sampled into a training set, a validation set, and a test set according to a preset ratio, for example, the preset ratio is 7:1:2. The training set is used to train the model data parameters, the validation set is used to observe the performance changes during model training to avoid underfitting or overfitting, and the test set is used to evaluate the model effect.

[0088] When the test set is used to test and evaluate the trained volatility prediction model, the evaluation indicators used are mainly the indicators of the regression model. The specific indicator calculation method can be any of the following methods:

[0089] (1) A = sum(abs(predicted value - actual value)) / sum(abs(actual value))

[0090] (2) B = avg (sum (abs (predicted value - actual value)))

[0091] (3) C = avg (abs (predicted value - actual value) / abs (actual value))

[0092] Among them, sum represents the sum operation, abs represents the absolute value operation, and avg represents the average value operation.

[0093] The smaller the value obtained by the calculation method of the above indicators (1)-(3), the higher the prediction accuracy of the trained model.

[0094] The fluctuation prediction model training method provided in the embodiments of the present disclosure obtains a fluctuation prediction model based on the historical operating data of the target object, historical operation information and the historical fluctuation amount of the key parameters after the intervention operation represented by the historical operation information is performed on the target object. This enables the user to determine the target operation information based on the fluctuation prediction model, achieves a relatively objective evaluation of the target object, and has a high prediction accuracy, reduces the amount of manual work, and improves processing efficiency.

[0095] Figure 6 FIG. 6 is a schematic diagram 600 of a second embodiment of a fluctuation prediction model training method according to the present disclosure. Figure 6 As shown, the method includes:

[0096] Step S601: Acquire historical operation data and historical operation information of a target object and historical fluctuations of key parameters after an intervention operation represented by the historical operation information is performed on the target object.

[0097] In this embodiment, step S601 is the same as step 501 in embodiment 500 and will not be described again here.

[0098] Step S602: Using historical operating data and historical operation information as input data for the first sub-model and the second sub-model, and using the first key parameter as expected output data, perform a first training on the first sub-model to obtain a first prediction sub-model; and using the second key parameter as expected output data, perform a second training on the second sub-training model to obtain a second prediction sub-model.

[0099] In this embodiment, the key parameters include a first key parameter and a second key parameter. Correspondingly, the initial causal inference model includes a first sub-model corresponding to the first key parameter and a second sub-model corresponding to the second key parameter. In some embodiments, the first key parameter and the second key parameter may be order quantity and revenue, respectively.

[0100] In this embodiment, when there are two key parameters, a multi-objective joint training method is used to train the fluctuation prediction model. Specifically, the historical operation data and historical operation information are used as input data of the first sub-model, and the first key parameter is used as the expected output data. The first sub-model is trained to obtain the first prediction sub-model. The historical operation data and historical operation information are used as input data of the second sub-model, and the second key parameter is used as the expected output data. The second sub-model is trained to obtain the second prediction sub-model. The fluctuation prediction model includes the trained first prediction sub-model and the second prediction sub-model.

[0101] This embodiment obtains a fluctuation prediction model by selecting multiple label data and performing multi-objective joint training, so that the trained model can evaluate the target object from multiple dimensions, thereby better meeting user needs.

[0102] The multi-objective joint training process in the disclosed embodiment uses a joint loss value to optimize model parameters. Specifically, during the first training and the second training, the parameters of the first sub-model and the second sub-model are tuned based on the joint loss value, respectively, wherein the joint loss value is determined by weighted summing the first loss value and the second loss value, the first loss value is the loss value calculated corresponding to the first sub-model, and the second loss value is the loss value calculated corresponding to the second sub-model.

[0103] Based on the same concept as the method for determining operation information based on causal inference provided by the present disclosure, an embodiment of the present disclosure also provides a device for determining operation information based on causal inference. Figure 7 FIG. 1 is a schematic diagram of a structure of an embodiment of an operation information determination device based on causal inference according to the present disclosure. Figure 7 As shown, the device comprises:

[0104] The information collection module 71 is configured to determine the operation data of the target object and at least one candidate operation information, where the candidate operation information is used to characterize the candidate intervention operation that can be implemented on the target object; the prediction module 72 is configured to input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain the key parameter fluctuation amount of the target object corresponding to each candidate operation information, and the fluctuation prediction model is trained based on the causal inference model; the determination module 73 is configured to determine the target operation information from at least one candidate operation information based on the key parameter fluctuation amount.

[0105] The operation information determination device based on causal inference provided in the embodiments of the present disclosure also has functional modules corresponding to the various method steps of the operation information determination method based on causal inference provided in the embodiments of the present disclosure. For the specific functions of each functional module and the relationship between them, please refer to the description of the various method steps of the operation information determination method based on causal inference provided in the embodiments of the present disclosure, which will not be repeated here.

[0106] The operation information determination device based on causal inference provided by the embodiment of the present disclosure predicts the fluctuation amount of the key parameters of the target object through a pre-trained fluctuation prediction model, and then determines the target operation information according to the predicted fluctuation amount, thereby achieving a relatively objective evaluation of the target object and having a high prediction accuracy, reducing the amount of manual work and improving processing efficiency.

[0107] Based on the same concept as the fluctuation prediction model training method provided in the present disclosure, an embodiment of the present disclosure also provides a fluctuation prediction model training device. Figure 8 FIG. 1 is a schematic diagram of a structure of an embodiment of a fluctuation prediction model training device according to the present disclosure. Figure 8 As shown, the device comprises:

[0108] The feature acquisition module 81 is configured to acquire the historical operation data, historical operation information and historical fluctuations of key parameters of the target object after the intervention operation represented by the historical operation information is performed on the target object; the training module 82 is configured to use the historical operation data and historical operation information as input data and the historical fluctuations as expected output data to train the initial training model to obtain a fluctuation prediction model.

[0109] The fluctuation prediction model training device provided in the embodiments of the present disclosure also has functional modules corresponding to the various method steps of the fluctuation prediction model training method provided in the embodiments of the present disclosure. For the specific functions of each functional module and the relationship between them, please refer to the description of the various method steps of the fluctuation prediction model training method provided in the embodiments of the present disclosure, which will not be repeated here.

[0110] The fluctuation prediction model training device provided in the embodiment of the present disclosure obtains a fluctuation prediction model based on the historical operating data of the target object, the historical operation information and the historical fluctuation amount of the key parameters after the intervention operation represented by the historical operation information is performed on the target object, thereby enabling the user to determine the target operation information based on the fluctuation prediction model, achieving a relatively objective evaluation of the target object, and having a high prediction accuracy, reducing the amount of manual work and improving processing efficiency.

[0111] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0112] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0113] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0114] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0115] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as an operation information determination method based on causal inference or a fluctuation prediction model training method. For example, in some embodiments, the method may be implemented as a computer software program based on a causal inference operation information determination method or a fluctuation prediction model training method, which is tangibly included in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 803 and executed by the computing unit 901, one or more steps of the operation information determination method based on causal inference or the fluctuation prediction model training method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured in any other appropriate manner (eg, by means of firmware) to execute an operation information determination method based on causal inference or a fluctuation prediction model training method.

[0116] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0117] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0118] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0119] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0120] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0121] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0122] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0123] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining operation information based on causal inference, comprising: Determining operation data of a target object and at least one candidate operation information, wherein the candidate operation information is used to characterize a candidate intervention operation that can be performed on the target object; Inputting the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain a fluctuation amount of a key parameter of the target object corresponding to each candidate operation information, wherein the fluctuation prediction model is trained based on a causal inference model; Based on the key parameter fluctuation amount, target operation information is determined from the at least one candidate operation information.

2. The method according to claim 1, wherein: The operational data includes mutable data and immutable data; The variable data includes at least one of the following: order data, waybill data, evaluation feedback data, attention frequency data and time feature data of the target object; The immutable data includes: portrait data of the target object.

3. The method according to claim 2, wherein: The step of inputting the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain the fluctuation amount of the key parameter of the target object corresponding to each candidate operation information includes: For any candidate operation information, concatenate the variable data, the immutable data and the candidate operation information to obtain target feature data; The target feature data is input into a pre-trained fluctuation prediction model to obtain the fluctuation amount of the key parameter of the target object corresponding to the candidate operation information.

4. The method according to any one of claims 1 to 3, further comprising: Acquire multiple operation data of the target object in multiple time periods; Based on the multiple operation data, the step of inputting the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model is performed multiple times to obtain fluctuation amounts of key parameters corresponding to each candidate operation information in the multiple time periods; The determining target operation information from the at least one candidate operation information based on the key parameter fluctuation amount includes: Determine the total amount of key parameter fluctuations corresponding to each candidate operation information based on the key parameter fluctuations corresponding to each candidate operation information in the multiple time periods; In response to the total amount of fluctuation of the key parameter satisfying a preset condition, the corresponding candidate operation information is determined as the target operation information.

5. The method according to any one of claims 1 to 4, wherein: The causal inference model includes a dual machine learning model.

6. A fluctuation prediction model training method, comprising: Acquire historical operation data and historical operation information of the target object and historical fluctuations of key parameters after the target object is subjected to an intervention operation represented by the historical operation information; The historical operating data and historical operation information are used as input data, and the historical fluctuation amount is used as expected output data, and the initial causal inference model is trained to obtain a fluctuation prediction model.

7. The method according to claim 6, wherein the key parameter comprises a first key parameter and a second key parameter; the initial causal inference model comprises a first sub-model corresponding to the first key parameter and a second sub-model corresponding to the second key parameter; The method uses the historical operation data and historical operation information as input data and the historical fluctuation amount as expected output data to train the initial training model to obtain a fluctuation prediction model, including: Using the historical operating data and historical operation information as input data for the first sub-model and the second sub-model; Taking the first key parameter as expected output data, performing a first training on the first sub-model to obtain a first prediction sub-model; and Taking the second key parameter as expected output data, performing a second training on the second sub-training model to obtain a second prediction sub-model; Wherein, the fluctuation prediction model includes the first prediction sub-model and the second prediction sub-model.

8. The method according to claim 7, wherein: During the first training and the second training, parameter tuning is performed based on a joint loss value, respectively. The joint loss value is determined by weighted summing the first loss value and the second loss value. The first loss value is a loss value corresponding to the first sub-model, and the second loss value is a loss value corresponding to the second sub-model.

9. The method according to claim 6, wherein: The step of acquiring the historical operation data and historical operation information of the target object and the historical fluctuation amount of the key parameters of the target object after the intervention operation represented by the historical operation information is performed includes: Determine sampling timestamp; Acquiring operation data within a preset time period before the sampling timestamp as the historical operation data; Determine the operation information corresponding to the intervention operation occurring at the sampling timestamp as the historical operation information; The fluctuation amount of the key parameter of the target object within a unit time after the sampling timestamp is determined as the historical fluctuation amount.

10. An operation information determination device based on causal inference, comprising: an information collection module configured to determine operation data of a target object and at least one candidate operation information, wherein the candidate operation information is used to characterize a candidate intervention operation that can be implemented on the target object; A prediction module is configured to input the operation data and at least one candidate operation information into a pre-trained fluctuation prediction model to obtain a fluctuation amount of a key parameter of the target object corresponding to each candidate operation information, wherein the fluctuation prediction model is trained based on a causal inference model; The determination module is configured to determine target operation information from the at least one candidate operation information based on the key parameter fluctuation amount.

11. A fluctuation prediction model training device, comprising: A feature acquisition module configured to acquire historical operation data and historical operation information of a target object and historical fluctuations of key parameters of the target object after an intervention operation represented by the historical operation information is performed on the target object; The training module is configured to use the historical operating data and historical operation information as input data and the historical fluctuation amount as expected output data to train the initial training model to obtain a fluctuation prediction model.

12. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method of any one of claims 1 to 5 or the method of claims 6 to 9.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 5 or the method according to claims 6 to 9.

14. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 5 or the method according to claims 6 to 9.

Citation Information

Patent Citations

  • Causal inference model training method and device, and information prompting method and device

    CN113112311A

  • Construction method and device of credit risk prediction model, equipment and storage medium

    CN116739606A

  • Model information determination method, device and system

    CN117216530A

  • Data driven customer churn analytics

    US20230206155A1