A method and device for determining a business processing strategy

Abstract

The present invention discloses a method and device for determining a business processing strategy, and relates to the field of computer technology. A specific implementation of the method includes: in response to a business processing request from a user, obtaining user portrait data and business status data of the user; serializing the user portrait data and business status data to obtain serialization features; inputting the serialization features into a pre-trained business processing model to obtain the output result of the business processing model; and determining the business processing strategy corresponding to the business processing request according to the output result. The method provided by the embodiment of the present invention enables the timing and continuity of data to be reflected through serialization processing, which facilitates the model to extract timing information and deep semantic information, thereby making business processing more efficient and accurate, and improving user experience.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for determining a business processing strategy. Background Art

[0002] When consumers shop on e-commerce platforms, transaction disputes usually arise, which can be resolved by compensation. Existing compensation review solutions have problems such as insufficient expression of data features and insufficient model interpretability. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a method and device for determining a business processing strategy, which can determine the business processing strategy by serializing user portrait data and business status data and inputting the data into a pre-trained business processing model, thereby improving the accuracy and interpretability of model predictions, thereby making business processing more efficient and accurate, and improving user experience.

[0004] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for determining a service processing strategy is provided, comprising:

[0005] In response to a user's business processing request, obtaining user portrait data and business status data of the user;

[0006] Serializing the user portrait data and the business status data to obtain serialization features;

[0007] Inputting the serialized features into a pre-trained business process model to obtain an output result of the business process model;

[0008] A service processing strategy corresponding to the service processing request is determined according to the output result.

[0009] Optionally, before inputting the serialized features into the pre-trained business processing model, the method further includes:

[0010] Constructing a training sample set and a test sample set, and pre-training a neural network model according to the training sample set to obtain a first network model;

[0011] Determine, according to the test sample set and the first network model, a first output result corresponding to each test sample in the test sample set;

[0012] The business processing model is trained according to the test sample set and the first output results corresponding to each test sample.

[0013] Optionally, training the business processing model according to the test sample set and the first output result corresponding to each test sample includes:

[0014] Inputting each test sample in the test sample set into a regression model to obtain a second output result corresponding to each test sample; the regression model includes each weight coefficient corresponding to each serialization feature;

[0015] Calculate the loss value corresponding to the test sample according to the second output result corresponding to the test sample, the first output result and the true result of the test sample;

[0016] The loss value is iteratively calculated according to the various weight coefficients, and when the increment of the loss value is less than a preset threshold, the various weight coefficients are determined to obtain the business processing model.

[0017] Optionally, calculating the loss value corresponding to the test sample according to the second output result corresponding to the test sample, the first output result and a true result in the test sample includes:

[0018] Calculating a loss function of the second output result and the first output result to determine a first loss value;

[0019] Calculating a loss function between the second output result and the true result to determine a second loss value;

[0020] The first loss value and the second loss value are weightedly summed to determine the loss value corresponding to the test sample.

[0021] Optionally, before serializing the user portrait data and the business status data, the method further includes:

[0022] The user portrait data and the business status data are discretized.

[0023] Optionally, the output result is a probability of adopting a preset first service processing strategy;

[0024] Determining a business processing strategy corresponding to the business processing request according to the output result includes: judging whether the probability is greater than or equal to a probability threshold; if so, using the first business processing strategy as the business processing strategy corresponding to the business processing request; otherwise, using a preset second business processing strategy as the business processing strategy corresponding to the business processing request.

[0025] Optionally, before adopting the preset second service processing strategy as the service processing strategy corresponding to the service processing request, the method further includes: confirming that no manual processing result for the service processing request has been received;

[0026] Otherwise, a business processing strategy corresponding to the business processing request is determined according to the manual processing result.

[0027] Optionally, after determining the business processing strategy corresponding to the business processing request according to the manual processing result, the method further includes: determining a correction coefficient corresponding to the business processing request according to the manual processing result, and updating the business processing model according to the correction coefficient.

[0028] According to another aspect of an embodiment of the present invention, there is provided a device for determining a service processing strategy, including:

[0029] An acquisition module, in response to a user's business processing request, acquires user portrait data and business status data of the user;

[0030] A data processing module performs serialization processing on the user portrait data and the business status data to obtain serialization features;

[0031] A model prediction module, inputting the serialized features into a pre-trained business processing model to obtain an output result of the business processing model;

[0032] A determination module determines a business processing strategy corresponding to the business processing request according to the output result.

[0033] According to another aspect of an embodiment of the present invention, there is provided an electronic device, including:

[0034] one or more processors;

[0035] a storage device for storing one or more programs,

[0036] When the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining a business processing strategy provided by the present invention.

[0037] According to another aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method for determining a service processing strategy provided by the present invention is implemented.

[0038] An embodiment of the above invention has the following advantages or beneficial effects: by serializing the user portrait data and business status data in the business processing request, serialization features are obtained, the serialization features are input into the pre-trained business processing model, the corresponding output results are obtained, and the business processing strategy corresponding to the business processing request is determined according to the output results. The method provided by the embodiment of the present invention characterizes semantic information and timing information by serializing the data, improves the current situation of poor accuracy and effect caused by insufficient model information, and realizes business processing more efficiently and accurately.

[0039] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific implementation examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0041] Figure 1 is a schematic diagram of the main process of a method for determining a service processing strategy according to an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the main process of another method for determining a service processing strategy according to an embodiment of the present invention;

[0043] Figure 3 is a flow chart of a method for determining a service processing strategy according to an implementation of an embodiment of the present invention;

[0044] Figure 4 is a flow chart of a method for determining a service processing strategy according to another implementation of an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of a method for determining a business processing model according to an embodiment of the present invention;

[0046] Figure 6 It is a flowchart of a method for determining a compensation review processing strategy according to an embodiment of the present invention;

[0047] Figure 7 is a schematic diagram of main modules of a device for determining a service processing strategy according to an embodiment of the present invention;

[0048] Figure 8 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0049] Fig. 9 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention 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 invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0051] Figure 1FIG. 1 is a schematic diagram of the main process of a method for determining a service processing strategy according to an embodiment of the present invention. Figure 1 As shown, the method for determining the business processing strategy is applied to the server, including:

[0052] Step S101: In response to a user's service processing request, obtaining the user's user portrait data and service status data;

[0053] Step S102: Serializing the user portrait data and the business status data to obtain serialization features;

[0054] Step S103: inputting the serialized features into the pre-trained business process model to obtain the output result of the business process model;

[0055] Step S104: Determine a service processing strategy corresponding to the service processing request according to the output result.

[0056] The method for determining a business processing strategy in an embodiment of the present invention can be applied to business processing involving compensation review in the after-sales field of an e-commerce platform.

[0057] In an embodiment of the present invention, the business processing request may be a submission request for a claim application form, and the claim application form may be transferred to the after-sales review link for review. When a business processing request is received, the user portrait data and business status data of the user are obtained. The user portrait data may include data such as user tag attributes, for example, it may include user after-sales preference tags such as red envelope compensation, refund, and return. The business status data may be data of each node or status information of the after-sales link (compensation link), including: at least one of the node information with state changes such as after-sales records, order status, logistics status, after-sales status, event id status, etc. The business status data represents the state evolution of the commodity in the transaction link, has a relatively strong temporal and logical expression, and represents the commodity semantic information of the commodity on the e-commerce platform.

[0058] The acquired user portrait data and business status data are original data. When used as model features for prediction, the user portrait data and business status data need to be converted in different ways according to the different requirements of data features of different models. For example, business status data needs to express the order of processing time or event order to express the temporal sequence and continuity of node or state information.

[0059] In an embodiment of the present invention, user portrait data and business status data are serialized and converted into serialized features to reflect the temporal sequence and continuity of node information, so as to enrich semantic features and temporal sequence information. The business status data is serialized, including the serialization of after-sales status data, logistics status data, order status data, and other business status data.

[0060] In an embodiment of the present invention, in step S102, before the user portrait data and the business status data are serialized, the process further includes: discretizing the user portrait data and the business status data. That is, the user portrait data and the business status data are discretized by using numbers, and different numbers can be used to represent whether a node has been executed or not, and different numbers can be used to represent whether a node has a certain label or not; for example, the number "0" can be used to represent that the node has not been executed, and the number "1" can be used to represent that the node has been executed; the number "0" is used to identify that a certain label is not present, and the number "1" is used to identify that a certain label is present. After the data is discretized and identified, it is serialized to obtain a serialization feature. When serializing, for the user portrait data, all the labels in the user portrait data of all users are used as a reference, and each label is used as an element in the serialization feature. When the user has a certain label or does not have a certain label, different numbers are used to represent, and the serialization feature is obtained; for the business status data, all the nodes or states of the business status data are used as a reference, and each node or state is used as an element in the serialization feature. When the node has been executed or not executed, different numbers are used to represent, and the serialization feature is obtained.

[0061] For example, user portrait data includes user label attributes, and labels corresponding to the user's preferred after-sales processing methods, such as "red envelope compensation", "repair", "return", and "exchange". If the label attribute is discretized into the number "1", and the absence of a certain label is represented by the number "0", when the user has two label attributes of "red envelope compensation" and "return", the user's user label attribute serialization feature is represented as [1,0,1,0].

[0062] For example, assuming that the after-sales status of a service order includes nodes such as "service order application", "pending review", "unified work platform review and collection", "customer service communication"... in order of events, the number "0" can be used to indicate that the node has not been executed, and the number "1" can be used to indicate that the node has been executed; then serialization processing is performed. For example, if the after-sales status of the service order has executed "pending review" but has not executed "same work platform review area", the serialization feature of the after-sales status of the service order can be represented as [1,1,0,0,...]. Then, the after-sales status of the service order can be known through the representation of the serialization feature, and the serialization feature can reflect the temporal and contextual relationship of the after-sales status of the service order.

[0063] The embodiment of the present invention serializes the business status data and user portrait data, and represents the business status data and user portrait data with serialized features, which can reflect the timing and continuity of each node or state in the business status data, record the state path and state change information, and make the data expression more reasonable. It is also convenient for subsequent business processing models to extract the deep semantics and timing information of the features, thereby improving the accuracy and effect of model prediction, and overcoming the deficiency that the traditional data feature expression method uses discrete representation and cannot obtain timing information.

[0064] In an embodiment of the present invention, after receiving a business processing request from a user, the method further includes: obtaining user information and product information, discretizing the user information and product information, obtaining discretized features, and inputting the discretized features into a pre-trained business processing model. That is, the features input into the pre-trained business processing model may include serialized features and discretized features. Among them, user information includes user attributes, such as the user's own attributes such as age and gender, and information such as the user's geographical location. In the after-sales field, it may also include the user's after-sales evaluation level (which can be obtained through customer service evaluation and scoring). Product information may include product attributes, such as product price, product size, etc. User information and product information are discretized, such as using the number "0" to represent the gender "male" and the number "1" to represent the gender "female", and are input into the business processing model as features, which can improve the accuracy of the business processing model prediction.

[0065] like Figure 2 As shown, in this embodiment of the present invention, in step S103, before the serialized features are input into the pre-trained business processing model, the following is also included:

[0066] Step S201: construct a training sample set and a test sample set, and pre-train the neural network model according to the training sample set to obtain a first network model;

[0067] Step S202: Determine a first output result corresponding to each test sample in the test sample set according to the test sample set and the first network model.

[0068] Step S203: training a business processing model according to the test sample set and the first output result corresponding to each test sample.

[0069] The interpretability of the model is an important indicator in the field of finance. Most of the existing claims review models are constructed using large-scale neural networks. The model effect is good, but the black box characteristics of the neural network make it difficult to know the internal working logic of the model, and the model is insufficiently interpretable. The model effect constructed by traditional machine learning is poor, but the interpretability is good. Therefore, the embodiment of the present invention combines the two to obtain a business processing model with good model effect and interpretability to perform business processing.

[0070] In an embodiment of the present invention, a business processing model is obtained by using knowledge distillation. First, a training sample set and a test sample set are constructed. The training sample set includes multiple training samples, and the test sample set includes multiple test samples. Both the training samples and the test samples include sample features and true results. The sample features include various features corresponding to the business processing request, including various serialized features and various discretized features. The true result is the result corresponding to the business processing request. For example, for the claim review business processing, the true result may be a pass or fail review, which may be represented by the numbers 1 and 0, respectively. The neural network model is pre-trained using the training sample set to determine the various parameters in the first network model, and then the first network model is obtained. The first network model is used as the teacher network, and the business processing model is used as the student network to train the business processing model. Among them, the neural network may be a deep neural network.

[0071] Optionally, after the training sample set is input into the neural network model, the neural network model is trained after embedding vector conversion, multi-head attention and fully connected layers to obtain a teacher network, and the test sample set is input into the trained teacher network to obtain the prediction results of each test sample, that is, the first output result. For example, the first output result may be the probability of passing and / or failing the business processing request.

[0072] In an embodiment of the present invention, a test sample in a test sample set is input into a first network model, and a first input result corresponding to the test sample can be obtained, and a business processing model can be trained through the test sample set and the first output result. Specifically, it includes: inputting each test sample in the test sample set into a regression model to obtain each second output result corresponding to each test sample; the regression model includes each weight coefficient corresponding to each sequence feature; according to the second output result corresponding to the test sample, the first output result and the true result of the test sample, the loss value corresponding to the test sample is calculated; according to each weight coefficient, the loss value is iteratively calculated, and when the increment of the loss value is less than a preset threshold, each weight coefficient is determined to obtain a business processing model.

[0073] The regression model in the embodiment of the present invention can be a logistic regression (LR) model. The LR model is a simple interpretable network model. The interpretability of the LR model is reflected in the fact that the feature weights completed through training represent the importance of each feature in the model, so that the sum of each feature multiplied by the corresponding feature weight (weight coefficient) is the output result.

[0074] In an embodiment of the present invention, a loss value corresponding to the test sample is calculated based on the second output result corresponding to the test sample, the first output result and the true result in the test sample, including: calculating the loss function of the second output result and the first output result to determine the first loss value; calculating the loss function of the second output result and the true result to determine the second loss value; and performing weighted summation of the first loss value and the second loss value to determine the loss value corresponding to the test sample.

[0075] In an embodiment of the present invention, the serialized features in the test sample set are used as regression input data, and the regression model is trained with the first output result and the true result as the target; the first loss value can be determined by calculating the cross entropy loss function between the second output result and the first output result; the second loss value can be determined by calculating the cross entropy loss function between the second output result and the true result. The first loss value and the second loss value can be directly added to obtain the loss value corresponding to the test sample, and the loss value is iteratively calculated. When the increment of the loss value is less than a preset threshold or no longer increases, the training of the regression model is stopped, thereby determining the various weight coefficients of the regression model, thereby determining the business processing model.

[0076] By training a regression model such as a linear regression model, we can obtain the weight coefficients corresponding to each serialization feature, that is, the proportion of each serialization feature, and achieve the interpretability of the business processing model. For example, if the weight coefficients corresponding to the after-sales status serialization feature, logistics status serialization feature, and order status serialization feature in the obtained business processing model are 0.1, 0.2, 0.3, and 0.4 respectively, it can be known that the order status serialization feature accounts for a large proportion in the business processing model.

[0077] The embodiment of the present invention provides an interpretable business processing model, which changes the existing business processing mode based on the black box model of the neural network, adopts the model interpretability mode of knowledge distillation, and compresses the complex neural network model into a simple network regression model such as the LR logistic regression model on the basis of basically unchanged or acceptable reduction of the model effect, so as to achieve the interpretability and easy online of the model, and effectively reduce the financial risk of business processing. It has been verified that the average accuracy of the logistic regression model prediction is 89.2%, and the average accuracy of the neural network model prediction is 91.3%. It is very worthwhile to exchange an accuracy of about 2.1% for an interpretable and easy-to-online model.

[0078] In an embodiment of the present invention, the output result is the probability of adopting a preset first business processing strategy; in step S104, the business processing strategy corresponding to the business processing request is determined based on the output result, including: judging whether the probability is greater than or equal to a probability threshold; if so, using the first business processing strategy as the business processing strategy corresponding to the business processing request; otherwise, adopting the preset second business processing strategy as the business processing strategy corresponding to the business processing request.

[0079] In an embodiment of the present invention, the preset first business processing strategy may be a business processing strategy for passing a business processing request, and the preset second business processing strategy may be a business processing strategy for failing a business processing request. The probability threshold may be a dynamic threshold, and the probability threshold may be obtained by means of data statistics, or the probability threshold may be adjusted dynamically according to business needs. For example, the business processing model audit is verified by random sampling, and when the model effect is slightly poor, the probability threshold may be increased to improve the accuracy of the model prediction.

[0080] In one implementation of the embodiment of the present invention, before the preset second business processing strategy is used as the business processing strategy corresponding to the business processing request, it also includes: confirming that no manual processing result for the business processing request has been received; otherwise, determining the business processing strategy corresponding to the business processing request according to the manual processing result. When the probability is less than the probability threshold, if no manual processing result for the business processing request has been received, the preset second business processing strategy is used as the business processing strategy corresponding to the business processing request; if the manual processing result for the business processing request is received, the business processing strategy for the business processing request is determined according to the manual processing result.

[0081] In the embodiment of the present invention, after determining the business processing strategy corresponding to the business processing request according to the manual processing result, the method further includes: determining the correction coefficient corresponding to the business processing request according to the manual processing result, and updating the business processing model according to the correction coefficient. The manual processing result can be an evaluation and scoring of the business processing request by the manual customer service, and the correction coefficient can be obtained after processing the evaluation and scoring result. For example, the correction coefficient can be obtained according to y coefficient =log(score+1) performs nonlinear processing on the evaluation score results, where y coefficient is the correction coefficient, score is the score obtained by the evaluation, and the product of the output result and the correction coefficient is used as the updated output result, that is, the probability multiplied by the correction coefficient is used as the updated probability, and the weight coefficient and other parameters in the business processing model are adjusted with the updated probability and the corresponding test sample to update the business processing model. The embodiment of the present invention performs feedback correction on the business processing model according to the correction coefficient to realize the correction of the business processing model, so as to further improve the accuracy of the business processing model.

[0082] like Figure 3 The figure shows a flow chart of determining a service processing strategy in an implementation mode of an embodiment of the present invention. The output result is the probability of adopting a preset first service processing strategy, and the service processing strategy corresponding to the service processing request is determined according to the output result, including:

[0083] Step S301: determine whether the probability is greater than or equal to the probability threshold, if yes, execute step S302; if no, execute step S303;

[0084] Step S302: using a preset first service processing policy as a service processing policy corresponding to the service processing request;

[0085] Step S303: determine whether the manual processing result indicates to adopt the preset first service processing strategy, if so, execute step S304, if not, execute step S305;

[0086] Step S304: using the preset first service processing policy as the service processing policy corresponding to the service processing request;

[0087] Step S305: Using the preset second service processing policy as the service processing policy corresponding to the service processing request.

[0088] like Figure 4 The flowchart of another implementation method of the present invention is shown. The output result is the probability of adopting the preset first business processing strategy, and the business processing strategy corresponding to the business processing request is determined according to the output result, including:

[0089] Step S401: determine whether the probability is greater than or equal to the probability threshold, if yes, execute step S402, if no, execute step S403;

[0090] Step S402: using a preset first service processing policy as a service processing policy corresponding to the service processing request;

[0091] Step S403: determining a correction coefficient corresponding to the business processing request according to the manual processing result, and determining an updated probability according to the correction coefficient;

[0092] Step S404: determine whether the updated probability is greater than or equal to the probability threshold; if yes, execute step S405; if not, execute step S406;

[0093] Step S405: using the preset first service processing policy as the service processing policy corresponding to the service processing request;

[0094] Step S406: Determine a service processing policy corresponding to the service processing request by using the preset second service processing policy.

[0095] like Figure 5 The figure shows a schematic diagram of a process for determining a business processing model. First, the acquired user portrait data and business status data are serialized to obtain serialized features. A training sample set and a test sample set are constructed based on the serialized features. The training sample set is input into the neural network. After embedding vector conversion, multi-head attention and full connection layer processing, the neural network is trained to obtain a teacher network. The test sample set is input into the teacher network to obtain a first output result (soft label) corresponding to each test sample, that is, a prediction result of the teacher network. The test sample set is input into the regression model as a training sample set, and the regression model is trained after embedding vector conversion. The second output results corresponding to each test sample can be obtained through the regression model. For each test sample, the loss value is calculated based on the second output result, the first output result and the true result (hard label). The loss value is calculated iteratively. When the increment of the loss value is less than the preset threshold, the training is stopped. After determining each parameter in the regression model, the weight coefficient corresponding to each serialized feature can be determined, and then the business processing model is obtained.

[0096] like Figure 6The figure is a flow chart of a method for determining a claim review and processing strategy provided by an embodiment of the present invention. In response to a claim application request from a user, the claim application request includes a claim slip, obtaining after-sales records, order status, logistics status, after-sales status, event ID status and other business status data and user portrait data corresponding to the claim slip, and performing serialization processing to obtain serialization features, obtaining user information and product information, and performing discretization processing to obtain discretization features, inputting the discretization features and serialization features into a business processing model, obtaining a probability of passing the review corresponding to the business processing request, and judging the probability Is it greater than or equal to the probability threshold? If so, the claim review is passed and the claim review process is completed; if not, it is transferred to manual review and processing, and the manual review determines whether the claim order is passed. If the manual review fails, the review opinion is given, and the user or relevant customer service re-applies for the claim order, and the claim order is returned to the claim order review pool; if the manual review of the claim application is passed, the manual customer service will manually evaluate and score the claim order, and the claim processing is completed. At the same time, the manual evaluation score is processed and used as a correction coefficient to update the probability of passing the review, so as to realize the feedback correction of the business processing model through the correction coefficient. Through after-sales data serialization and knowledge distillation, a closed-loop, self-correcting, and explainable claim review processing strategy is realized, making the claim review results more objective and fair, improving the efficiency of claim review, improving user experience, and reducing unnecessary claim losses of e-commerce companies.

[0097] The method provided by the embodiment of the present invention serializes the user portrait data and the business status data to obtain serialization features, and inputs the serialization features into the pre-trained business processing model to obtain output results, and obtains the business processing strategy corresponding to the business processing request according to the output results. By serializing the data to express the timing and continuity of the node information, the business processing model can extract the timing information and deep semantic information of each node, improving the current situation of low accuracy and poor effect caused by insufficient model information; the business processing model is obtained by guiding the training of the first network model in the form of knowledge distillation, which improves the model effect of the business processing model and has a certain degree of interpretability. At the same time, the self-correction and closed-loop capabilities of the business processing are realized by manual processing, making the business processing more efficient and accurate, thereby improving the user experience and reducing the financial risks of the business processing.

[0098] Figure 7 is a schematic diagram of main modules of an apparatus 700 for determining a service processing strategy according to an embodiment of the present invention, such as Figure 5 As shown, the device 700 for determining a service processing strategy includes:

[0099] The acquisition module 701 acquires the user's user portrait data and business status data in response to the user's business processing request;

[0100] The data processing module 702 performs serialization processing on the user portrait data and the business status data to obtain serialization features;

[0101] Model prediction module 703, inputting serialized features into a pre-trained business process model to obtain an output result of the business process model;

[0102] The determination module 704 determines the service processing strategy corresponding to the service processing request according to the output result.

[0103] In an embodiment of the present invention, the device also includes a data storage module, and the acquisition module 701 can obtain the user portrait data and business status data of the user from the data storage module.

[0104] In an embodiment of the present invention, the model prediction module 703 is further used to: construct a training sample set and a test sample set before inputting the serialized features into the pre-trained business processing model, pre-train the neural network model according to the training sample set, and obtain a first network model; determine the first output result corresponding to each test sample in the test sample set according to the test sample set and the first network model, and train the business processing model according to the test sample set and the first output result corresponding to each test sample.

[0105] In an embodiment of the present invention, the model prediction module 703 is further used to: input each test sample in the test sample set into the regression model to obtain a second output result corresponding to each test sample; the regression model includes each weight coefficient corresponding to each sequence feature; according to the second output result corresponding to the test sample, the first output result and the true result of the test sample, the loss value corresponding to the test sample is calculated; the loss value is iteratively calculated according to each weight coefficient, and when the increment of the loss value is less than a preset threshold, each weight coefficient is determined, and the regression model is determined as a business processing model.

[0106] In an embodiment of the present invention, the model prediction module 703 is further used to: calculate the loss function of the second output result and the first output result to determine the first loss value; calculate the loss function of the second output result and the true result to determine the second loss value; perform weighted summation of the first loss value and the second loss value to determine the loss value corresponding to the test sample.

[0107] In the embodiment of the present invention, the data processing module 702 is further used to discretize the user portrait data and the business status data before serializing them.

[0108] In an embodiment of the present invention, the output result is the probability of adopting a preset first business processing strategy; the determination module 704 is further used to: determine whether the probability is greater than or equal to a probability threshold; if so, use the first business processing strategy as the business processing strategy corresponding to the business processing request; otherwise, use the preset second business processing strategy as the business processing strategy corresponding to the business processing request.

[0109] In an embodiment of the present invention, the determination module 704 is also used to: before adopting the preset second business processing strategy as the business processing strategy corresponding to the business processing request, confirm that no manual processing result for the business processing request has been received; otherwise, determine the business processing strategy corresponding to the business processing request based on the manual processing result.

[0110] In the embodiment of the present invention, the determination module 704 is further used to: after determining the business processing strategy corresponding to the business processing request according to the manual processing result, determine the correction coefficient corresponding to the business processing request according to the manual processing result, and update the business processing model according to the correction coefficient. The determination module can also be called a threshold judgment and evaluation feedback module.

[0111] An embodiment of the present invention also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining the business processing strategy of the embodiment of the present invention.

[0112] The embodiment of the present invention further provides a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the method for determining a service processing strategy according to the embodiment of the present invention is implemented.

[0113] Figure 8 An exemplary system architecture 800 is shown to which the method for determining a service processing policy or the apparatus for determining service processing according to the embodiments of the present invention can be applied.

[0114] like Figure 8 As shown, system architecture 800 may include terminal devices 801, 802, 803, a network 804, and a server 805. Network 804 is used to provide a medium for communication links between terminal devices 801, 802, 803 and server 805. Network 804 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0115] Users can use terminal devices 801, 802, and 803 to interact with server 805 through network 804 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 801, 802, and 803, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0116] The terminal devices 801 , 802 , and 803 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0117] The server 805 may be a server that provides various services, such as a backend management server (only an example) that provides support for shopping websites browsed by users using the terminal devices 801, 802, and 803. The backend management server may analyze and process the received data such as product information query requests, and feed back the processing results (such as target push information, product information - only an example) to the terminal device.

[0118] It should be noted that the method for determining a service processing strategy provided in the embodiment of the present invention is generally executed by the server 805 , and accordingly, the device for determining a service processing strategy is generally disposed in the server 805 .

[0119] It should be understood that Figure 8 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0120] Reference below Fig. 9 , which shows a schematic diagram of the structure of a computer system 900 of a terminal device suitable for implementing an embodiment of the present invention. Fig. 9 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0121] like Fig. 9 As shown, the computer system 900 includes a central processing unit (CPU) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the system 900 are also stored. The CPU 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.

[0122] The following components are connected to the I / O interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage section 908 as needed.

[0123] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the central processing unit (CPU) 901, the above-mentioned functions defined in the system of the present invention are executed.

[0124] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, 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 above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0125] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0126] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes an acquisition module, a data processing module, a model prediction module, and a determination module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the determination module may also be described as a "module for determining a business processing strategy corresponding to a business processing request based on an output result."

[0127] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: in response to a user's business processing request, obtaining the user's user portrait data and business status data; serializing the user portrait data and business status data to obtain serialized features; inputting the serialized features into a pre-trained business processing model to obtain the output result of the business processing model; and determining the business processing strategy corresponding to the business processing request according to the output result.

[0128] According to the technical solution of the embodiment of the present invention, by serializing the user portrait data and the business status data, serialization features are obtained, and the serialization features are input into the pre-trained business processing model to obtain output results, and the business processing strategy corresponding to the business processing request is obtained according to the output results. By serializing the data to express the timing and continuity of the node information, the business processing model can extract the timing information and deep semantic information of each node, improving the current situation of low accuracy and poor effect caused by insufficient model information; and the business processing model is obtained by guiding the training of the first network model in the way of knowledge distillation, which improves the model effect of the business processing model and has a certain degree of interpretability. At the same time, the self-correction and closed-loop capabilities of the business processing are realized by manual processing, making the business processing more efficient and accurate, thereby improving the user experience and reducing the financial risks of the business processing.

[0129] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for determining a business processing strategy, It is characterized in that include: In response to a user's business processing request, obtaining user portrait data and business status data of the user; The service status data includes node information with status changes, and the node information includes each node of the service status data; Serializing the user portrait data and the business status data to obtain serialization features; Inputting the serialized features into a pre-trained business process model to obtain an output result of the business process model; Determine a business processing strategy corresponding to the business processing request according to the output result; The business status data is serialized to obtain a serialization feature, including: taking all nodes of the business status data as a reference, each node as an element in the serialization feature, and obtaining the serialization feature according to the executed or unexecuted state indicated by each node.

2. The method according to claim 1, It is characterized in that Before inputting the serialized features into the pre-trained business processing model, the method further includes: Constructing a training sample set and a test sample set, and pre-training a neural network model according to the training sample set to obtain a first network model; Determine, according to the test sample set and the first network model, a first output result corresponding to each test sample in the test sample set; The business processing model is trained according to the test sample set and the first output results corresponding to each test sample.

3. The method according to claim 2, It is characterized in that Training the business processing model according to the test sample set and the first output results corresponding to each test sample includes: Inputting each test sample in the test sample set into a regression model to obtain a second output result corresponding to each test sample; the regression model includes each weight coefficient corresponding to each serialization feature; Calculate the loss value corresponding to the test sample according to the second output result corresponding to the test sample, the first output result and the true result of the test sample; The loss value is iteratively calculated according to the various weight coefficients, and when the increment of the loss value is less than a preset threshold, the various weight coefficients are determined to obtain the business processing model.

4. The method according to claim 3, It is characterized in that Calculating the loss value corresponding to the test sample according to the second output result corresponding to the training sample, the first output result, and the true result in the test sample, includes: Calculating a loss function of the second output result and the first output result to determine a first loss value; Calculating a loss function between the second output result and the true result to determine a second loss value; The first loss value and the second loss value are weightedly summed to determine the loss value corresponding to the test sample.

5. The method according to claim 1, It is characterized in that Before serializing the user portrait data and the business status data, the method further includes: The user portrait data and the business status data are discretized.

6. The method according to claim 1, It is characterized in that The output result is the probability of adopting the preset first service processing strategy; Determining a business processing strategy corresponding to the business processing request according to the output result includes: judging whether the probability is greater than or equal to a probability threshold; if so, using the first business processing strategy as the business processing strategy corresponding to the business processing request; otherwise, using a preset second business processing strategy as the business processing strategy corresponding to the business processing request.

7. The method according to claim 6, It is characterized in that Before adopting the preset second service processing strategy as the service processing strategy corresponding to the service processing request, the method further includes: confirming that no manual processing result for the service processing request has been received; Otherwise, a business processing strategy corresponding to the business processing request is determined according to the manual processing result.

8. The method according to claim 7, It is characterized in that After determining the business processing strategy corresponding to the business processing request according to the manual processing result, the method further includes: determining a correction coefficient corresponding to the business processing request according to the manual processing result, and updating the business processing model according to the correction coefficient.

9. A device for determining a business processing strategy, It is characterized in that include: An acquisition module, in response to a user's business processing request, acquires user portrait data and business status data of the user; The service status data includes node information with status changes, and the node information includes each node of the service status data; A data processing module performs serialization processing on the user portrait data and the business status data to obtain serialization features; A model prediction module, inputting the serialized features into a pre-trained business processing model to obtain an output result of the business processing model; A determination module, which determines a business processing strategy corresponding to the business processing request according to the output result; Wherein, the data processing module is further used to: take all nodes of the business status data as a reference, each node as an element in the serialization feature, and obtain the serialization feature according to the executed or unexecuted state indicated by each node.

10. An electronic device, It is characterized in that include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

11. A computer readable medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images