Task allocation processing method and device

Abstract

The embodiment of the invention provides a task allocation processing method and device. The method comprises the steps of obtaining a to-be-allocated target task, and determining a target customer; customer feature information of a target customer and personnel feature information of each worker are obtained, and normalization processing is performed on each numerical feature information of the target customer and each worker; inputting each category feature information of the target customer and each worker into an embedded layer of a task completion rate prediction model to obtain a category feature vector; inputting the numerical value feature vector and the category feature vector of the target customer and each worker into a task completion rate prediction model to output a prediction completion rate of the target task corresponding to the worker; and sending the target task to a terminal corresponding to the target worker with the highest prediction completion rate. Based on the method provided by the invention, the target task can be ensured to be completed at the maximum probability, and the purpose of improving the task completion rate is achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a task allocation processing method and device. Background Art

[0002] The insurance sales management system can design a specific sales process based on the sales expectations of insurance products, and by clarifying the sales tasks at different time points in the sales process, assign corresponding sales tasks to staff members so that the staff members can complete the sales tasks and ultimately complete the sales of insurance products.

[0003] At present, the insurance sales management system has three ways of allocating sales tasks: allocation according to the generation time of the sales task, allocation according to the average workload, allocation according to the previous person in charge, etc.

[0004] However, the current way of allocating sales tasks in the insurance sales management system can only satisfy the normal execution of sales tasks by staff, but the completion rate of sales tasks for insurance products is generally low. Summary of the invention

[0005] The embodiments of the present application provide a task allocation processing method and device for improving the completion rate of insurance product sales tasks.

[0006] In a first aspect, an embodiment of the present application provides a task allocation processing method, which is applied to a server, and the method includes:

[0007] Obtaining the target tasks to be assigned, and determining the target customers corresponding to the target tasks;

[0008] Acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer;

[0009] Acquire personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member;

[0010] Normalizing each numerical feature information of the target customer to obtain a numerical feature vector of the target customer;

[0011] Input each category feature information of the target customer into a task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model:

[0012] Input each category feature information into the embedding layer to output the category feature vector of the target customer;

[0013] Inputting the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer;

[0014] For each staff member, normalize each numerical feature information of the staff member to obtain a numerical feature vector of the staff member;

[0015] For each staff member, input each category feature information of the staff member into the embedding layer to output the category feature vector of the staff member;

[0016] For each staff member, the numerical feature vector and the category feature vector of the staff member are input into the connection layer to output a complete feature vector corresponding to the staff member;

[0017] For each staff member, the complete feature vector corresponding to the staff member and the complete feature vector corresponding to the target customer are input into the first module for fusion and activation processing, so as to output the first fusion feature corresponding to the staff member;

[0018] Selecting a reference worker from each worker, and determining a complete feature vector corresponding to the reference worker as a reference complete feature vector;

[0019] For each staff member, the complete feature vector corresponding to the staff member and the reference complete feature vector are input into the second module for linear transformation processing and dimensionality reduction processing, so as to output the second fusion feature corresponding to the staff member;

[0020] For each worker, the first fusion feature and the second fusion feature corresponding to the worker are input into the output layer for prediction processing, so as to output the predicted completion rate of the target task corresponding to the worker;

[0021] Comparing the predicted completion rates of the target tasks for each worker, and determining the worker with the highest predicted completion rate as the target worker;

[0022] The target task is sent to the terminal corresponding to the target worker.

[0023] In a possible implementation manner, normalizing each numerical feature information of the target customer to obtain the numerical feature vector of the target customer includes:

[0024] Obtaining a preset upper limit value and a preset lower limit value respectively corresponding to each numerical characteristic information of the target customer;

[0025] Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information;

[0026] splicing the ratio results corresponding to each numerical feature information of the target customer to obtain the numerical feature vector of the target customer;

[0027] The normalizing each numerical feature information of the staff member to obtain the numerical feature vector of the staff member includes:

[0028] Obtaining a preset upper limit value and a preset lower limit value corresponding to each numerical characteristic information of the staff member;

[0029] Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information;

[0030] The ratio results corresponding to each numerical feature information of the staff member are spliced ​​to obtain the numerical feature vector of the staff member.

[0031] In a possible implementation, the step of screening reference workers from among the workers includes:

[0032] Obtaining a task completion record, wherein the task completion record includes each completed historical task and the staff member who completed each historical task;

[0033] The worker who has completed the largest number of historical tasks is determined as the reference worker.

[0034] In one possible implementation, the embedding layer includes an autoencoder fully connected neural network model;

[0035] Accordingly, the step of inputting each category feature information into the embedding layer to output the category feature vector of the target customer includes:

[0036] Input each category feature information into the autoencoder fully connected neural network model to output the category feature vector of the target customer;

[0037] Accordingly, the step of inputting each category feature information of the staff member into the embedding layer to output the category feature vector of the staff member includes:

[0038] Each category feature information of the staff member is input into the autoencoder fully connected neural network model to output the category feature vector of the staff member.

[0039] In a possible implementation, the training process of the autoencoder fully connected neural network model includes:

[0040] Acquire first training data, where the first training data includes training category feature information;

[0041] Inputting the training category feature information into the autoencoder fully connected neural network model to output an iteration result;

[0042] Inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result;

[0043] Circularly executing the process of inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result;

[0044] A loss function is used to calculate the loss value between each input and the corresponding output of the autoencoder fully connected neural network model. If the loss value exceeds a first preset value, the parameters of the autoencoder fully connected neural network model are updated until the loss value is less than the first preset value, and the training of the autoencoder fully connected neural network model is stopped.

[0045] In a possible implementation, the training process of the task completion rate prediction model includes:

[0046] Acquire multiple interaction records, wherein the interaction records include customer information, staff information, and interaction content between the customer and the staff;

[0047] Based on the interaction content between the customer and the staff in each interaction record, determine the actual completion rate corresponding to each interaction record;

[0048] Acquire second training data, wherein the second training data includes customer feature information of a customer and staff feature information of a staff member corresponding to each interaction record, wherein the customer feature information includes each numerical feature information and each category feature information of the customer, and the staff feature information includes each numerical feature information and each category feature information of the staff member;

[0049] Initializing model parameters of the task completion rate prediction model;

[0050] Input each numerical feature information and each category feature information of the customer and each numerical feature information and each category feature information of each staff member into the task completion rate prediction model to output the predicted completion rate corresponding to each staff member;

[0051] The loss function is used to calculate the loss value between the predicted completion rate and the actual completion rate of each worker.

[0052] If the loss value exceeds a second preset value, the model parameters of the task completion rate prediction model are updated until the second training data completes the training of the task completion rate prediction model.

[0053] In a second aspect, an embodiment of the present application provides a task allocation processing device, the device comprising:

[0054] An acquisition module is used to acquire the target task to be assigned and determine the target customer corresponding to the target task;

[0055] The acquisition module is further used to acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer;

[0056] The acquisition module is further used to acquire the personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member;

[0057] A processing module, used for normalizing each numerical feature information of the target customer to obtain a numerical feature vector of the target customer;

[0058] The input module is used to input each category feature information of the target customer into the task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model:

[0059] The input module is further used to input each category feature information into the embedding layer to output the category feature vector of the target customer;

[0060] The input module is further used to input the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer;

[0061] The processing module is further used to normalize each numerical feature information of each staff member to obtain a numerical feature vector of the staff member;

[0062] The input module is further used to input each category feature information of each staff member into the embedding layer to output the category feature vector of the staff member;

[0063] The input module is further used to input the numerical feature vector and the category feature vector of each staff member into the connection layer to output a complete feature vector corresponding to the staff member;

[0064] The input module is further used to input, for each staff member, the complete feature vector corresponding to the staff member and the complete feature vector corresponding to the target customer into the first module for fusion and activation processing, so as to output a first fusion feature corresponding to the staff member;

[0065] A determination module, used for selecting a reference worker from each worker, and determining a complete feature vector corresponding to the reference worker as a reference complete feature vector;

[0066] The input module is further used to input the complete feature vector corresponding to each staff member and the reference complete feature vector into the second module for linear transformation processing and dimensionality reduction processing, so as to output the second fusion feature corresponding to the staff member;

[0067] The input module is further used to input the first fusion feature and the second fusion feature corresponding to each staff member into the output layer for prediction processing, so as to output the predicted completion rate of the target task corresponding to the staff member;

[0068] The determination module is further used to compare the predicted completion rates of the target tasks corresponding to each worker, and determine the worker with the highest predicted completion rate as the target worker;

[0069] The sending module is used to send the target task to the terminal corresponding to the target staff member.

[0070] In a third aspect, an embodiment of the present application provides a computer device, including: a memory, a processor;

[0071] Memory stores computer-executable instructions;

[0072] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0073] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0074] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0075] The task allocation processing method and device provided in the embodiment of the present application normalize each numerical feature information of the target customer and each staff member to obtain the numerical feature vector of the target customer and each staff member, input each category feature information of the target customer and each staff member into the embedding layer of the task completion rate prediction model to output the category feature vector of the target customer and each staff member; input the numerical feature vector and the category feature vector of the target customer and each staff member into the connection layer of the task completion rate prediction model to output the complete feature vector corresponding to the target customer and each staff member; input the complete feature vector corresponding to each staff member and the target customer into the first module of the task completion rate prediction model Perform fusion and activation processing to output the first fusion feature corresponding to each staff member; select reference staff members from each staff member, and determine the complete feature vector corresponding to the reference staff member as the reference complete feature vector; input each staff member and the reference complete feature vector into the second module of the task completion rate prediction model for linear transformation processing and dimensionality reduction processing to output the second fusion feature corresponding to each staff member; input the first fusion feature and the second fusion feature corresponding to each staff member into the output layer of the task completion rate prediction model for prediction processing to output the predicted completion rate of the target task corresponding to the staff member; send the target task to the terminal corresponding to the target staff member with the highest predicted completion rate. This application can ensure that the target task can be completed with the highest probability by predicting the completion rate of each staff member for the target task, thereby achieving the purpose of improving the task completion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0077] Figure 1 A schematic diagram of a scenario for this application example;

[0078] Figure 2 Schematic diagram of the task allocation processing method provided in this application Figure 1 ;

[0079] Figure 3 This is a schematic diagram of the structure of the task completion rate prediction model for the example;

[0080] Figure 4 Schematic diagram of the structure of the cross-product transformation layer of the example;

[0081] Figure 5 Schematic diagram of the task allocation processing method provided in this application Figure 2 ;

[0082] Figure 6 Schematic diagram of the task allocation processing method provided in this application Figure 3 ;

[0083] Figure 7 Schematic diagram of the task allocation processing method provided in this application Figure 4 ;

[0084] Figure 8 Schematic diagram of the task allocation processing method provided in this application Figure 5 ;

[0085] Fig. 9 The schematic diagram of the structure of the fully connected neural network model of the autoencoder is an example;

[0086] Fig.10 Schematic diagram of the task allocation processing method provided in this application Figure 6 ;

[0087] Fig.11 Schematic diagram of the task allocation processing method provided in this application Figure 7 ;

[0088] Fig.12 A schematic diagram of the structure of the task allocation processing device provided by this application;

[0089] Fig.13 A schematic diagram of the structure of the computer device provided for this application.

[0090] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0091] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0092] Figure 1 A schematic diagram of a scenario for this application example, such as Figure 1As shown, the specific application scenarios of this application mainly include the server ① of the insurance sales management system and the client ② of the staff. The server ① can design a specific sales process according to the sales expectations of the insurance products, and by clarifying the sales tasks at different time points in the sales process, send the sales tasks to the client ② of the staff to assign the corresponding sales tasks to the staff, so that the staff can complete the sales tasks and finally complete the sales of the insurance products.

[0093] In combination with the above scenarios, it can be seen that in the prior art, the insurance sales management system has the following ways of allocating sales tasks: allocating by the time when the sales task is generated, allocating by the average workload, allocating by the last person in charge, etc. However, the current insurance sales management system can only satisfy the normal execution of sales tasks by the staff, but the completion rate of insurance product sales tasks is generally low.

[0094] The task assignment processing method provided by the present application first normalizes each numerical feature information of the target customer and each staff member to obtain the numerical feature vector of the target customer and each staff member, and then inputs each category feature information of the target customer and each staff member into the embedding layer of the task completion rate prediction model to output the category feature vector of the target customer and each staff member; then inputs the numerical feature vector and the category feature vector of the target customer and each staff member into the connection layer of the task completion rate prediction model to output the complete feature vector corresponding to the target customer and each staff member; then inputs the complete feature vector corresponding to each staff member and the target customer into the first module of the task completion rate prediction model for integration. The first fusion feature corresponding to each staff member is output through the combination and activation process; then the reference staff member is selected from each staff member, and the complete feature vector corresponding to the reference staff member is determined as the reference complete feature vector; then each staff member and the reference complete feature vector are input into the second module of the task completion rate prediction model for linear transformation and dimensionality reduction processing to output the second fusion feature corresponding to each staff member; then the first fusion feature and the second fusion feature corresponding to each staff member are input into the output layer of the task completion rate prediction model for prediction processing to output the predicted completion rate of the target task corresponding to the staff member; finally, the target task is sent to the terminal corresponding to the target staff member with the highest predicted completion rate. This application can ensure that the target task can be completed with the highest probability by predicting the completion rate of each staff member for the target task, thereby achieving the purpose of improving the task completion rate.

[0095] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and corresponding operation entrances shall be provided for users to choose to authorize or refuse.

[0096] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0097] Figure 2 Schematic diagram of the task allocation processing method provided in this application Figure 1 ,like Figure 2 As shown, the method includes:

[0098] S201, obtaining a target task to be assigned, and determining a target customer corresponding to the target task.

[0099] The execution subject of this embodiment is Figure 1 The server in .

[0100] Combined with the scenario example, a dedicated sales task can be created for each customer to be promoted. For example, if there are multiple customers to be promoted, such as user A, user B, and user C, a sales task corresponding to each customer to be promoted, such as user A, user B, and user C, can be created. The target task is the next sales task to be executed, and the customer corresponding to the target task is the target customer. For example, if the sales task for user A is the target task, then user A is the target customer.

[0101] S202: Acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer.

[0102] Combined with the scenario example, the numerical feature information includes but is not limited to: age, family income, number of properties, number of vehicles, assets, liabilities, weight, height, current premiums paid and insurance coverage, etc. Category feature information includes but is not limited to: occupation, nationality, work unit, residential address, education level, eating habits, exercise habits, gender (male is 1, female is 0), family role-economic pillar (yes is 1, no is 0), family role-need to support parents (need is 1, no is 0), property type-self-occupied housing (yes is 1, no is 0), vehicle type-private car (yes is 1, no is 0), whether suffering from disease (yes is 1, no is 0), whether there is a history of surgery (yes is 1, no is 0), whether smoking (yes is 1, no is 0), currently held insurance type-life insurance (yes is 1, no is 0) and currently held insurance type-medical insurance (yes is 1, no is 0), etc.

[0103] S203: Acquire personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member.

[0104] In combination with the scenario example, the obtained numerical feature information of each staff member is of the same type as the numerical feature information of the target customer, and the obtained category feature information of each staff member is of the same type as the category feature information of the target customer.

[0105] S204: normalize each numerical feature information of the target customer to obtain a numerical feature vector of the target customer.

[0106] Combined with the scenario example, each numerical feature information of the target customer is normalized to avoid excessive attention caused by the absolute value of a certain numerical feature information being too large. Then the results of the normalization of each numerical feature information are spliced ​​in a preset order to obtain the numerical feature vector of the target customer.

[0107] S205, inputting each category feature information of the target customer into a task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model:

[0108] Combined with the scene example, Figure 3 The structural diagram of the task completion rate prediction model is shown in the following figure: Figure 3As shown in the figure, the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer. The first module includes three fully connected layers, each of which is a deep neural network with ReLU as the activation function, and the dimensions are: 1024, 512 and 256 respectively. The second module includes a cross product transformation layer and a weight layer, and the output layer includes a Sigmoid activation function.

[0109] S206: Input each category feature information into the embedding layer to output the category feature vector of the target customer.

[0110] Combined with the scenario example, the category features of target customers are divided into binary features and non-binary features. The binary features are gender (male is 1, female is 0), family role-economic pillar (yes is 1, no is 0), family role-need to support parents (need is 1, no is 0), property type-self-occupied housing (yes is 1, no is 0), vehicle type-private car (yes is 1, no is 0), whether suffering from disease (yes is 1, no is 0), whether there is a history of surgery (yes is 1, no is 0), whether smoking (yes is 1, no is 0), current insurance type-life insurance (yes is 1, no is 0) and current insurance type-medical insurance (yes is 1, no is 0) and other numerical features, and there are only two values ​​​​0 and 1. For binary features, since they are already numerical features, they do not need to be encoded through the embedding layer. For non-binary features, such as occupation, nationality, work unit, residential address, education level, eating habits and exercise habits, all non-binary features can be spliced ​​into a temporary vector in a preset order, input into the embedding layer, encoded by the embedding layer, and then output a 20-dimensional encoding vector. In this example, the 20-dimensional encoding vector output by the embedding layer is used as the category feature vector of the target user. The encoding method of the encoder's fully connected layer for each non-binary feature can be one-hot encoding. Taking occupation as an example, in the embedding layer, occupations can be set to 10 types, namely: occupation 1, occupation 2...occupation 10. If the target customer's occupation is occupation 1, the target customer's occupation after one-hot encoding by the embedding layer is (1, 0, 0, 0, 0, 0, 0, 0, 0). Similarly, for the residential address, in the embedding layer, the residential address can be divided into N types, namely residential address 1, residential address 2... residential address N. If the residential address of the target customer is residential address 1, then the residential address of the target customer is (0, 1, 0, 0, 0...) after one-hot encoding in the embedding layer. In this way, the encoding of the temporary vector is completed through the embedding layer, and the encoding result of the temporary vector is concatenated with each binary feature to obtain the category feature vector of the target customer.

[0111] S207 , inputting the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer.

[0112] Combined with the scene example, combined with Figure 3 , the numerical feature vector and the category feature vector of the target customer are input into the connection layer for concatenation to obtain the complete feature vector of the target customer.

[0113] S208. For each staff member, normalize each numerical feature information of the staff member to obtain a numerical feature vector of the staff member.

[0114] Combined with the scenario example, the processing method for the numerical features of each staff member is consistent with the processing method for the numerical features of the target customers. Specifically, for each staff member, each corresponding numerical feature information is normalized to avoid excessive attention caused by the absolute value of a certain numerical feature information being too large, and then the results of the normalization of each numerical feature information are spliced ​​in a preset order to obtain the numerical feature vector corresponding to the staff member.

[0115] S209: For each staff member, input each category feature information of the staff member into the embedding layer to output the category feature vector of the staff member.

[0116] Combined with the scenario example, the processing method for the category feature information of each staff member is consistent with the processing method for the category feature information of the target customer. Specifically, for each staff member, the category feature of the staff member is divided into binary features and non-binary features. Each non-binary feature of the staff member is first concatenated into a temporary vector, and then the temporary vector is input into the embedding layer to complete the encoding of the temporary vector. The encoding result of the temporary vector is concatenated with each binary feature to obtain the category feature vector of the staff member.

[0117] S2010. For each staff member, the numerical feature vector and the category feature vector of the staff member are input into the connection layer to output a complete feature vector corresponding to the staff member.

[0118] Combined with the scene example, combined with Figure 3 ,For each staff member, the staff member’s numerical feature vector and category feature vector are input into the connection layer for concatenation to obtain the staff member’s complete feature vector.

[0119] S2011. For each staff member, the complete feature vector corresponding to the staff member and the complete feature vector corresponding to the target customer are input into the first module for fusion and activation processing to output the first fusion feature corresponding to the staff member.

[0120] Combined with the scenario example, the complete feature vector of the target customer and the complete feature vector of the staff are connected in the connection layer and then input into the first module. The first module includes three layers of deep neural network with ReLU as the activation function. The dimensions are 1024, 512 and 256 respectively. Therefore, after being processed by the first module, the first fused feature with a dimension of 256 corresponding to the staff can be obtained.

[0121] S2012. Select a reference worker from among the workers, and determine the complete feature vector corresponding to the reference worker as the reference complete feature vector.

[0122] Combined with the scenario example, the reference staff member can be the staff member who has completed the most sales tasks among all staff members. For example, if staff member A is the staff member who has completed the most sales tasks, staff member A can be used as the reference staff member, and the complete feature vector of staff member A can be used as the reference complete feature vector.

[0123] S2013. For each staff member, the complete feature vector corresponding to the staff member and the reference complete feature vector are input into the second module for linear transformation and dimensionality reduction processing to output the second fusion feature corresponding to the staff member.

[0124] Combined with the scene example, combined with Figure 3 ,The second module includes a cross product transformation layer and a weight layer. For each worker, the obtained complete feature vector and the reference complete feature vector are input into the second module, and the cross and transformation layer performs linear transformation on the worker’s complete feature vector and the reference complete feature vector. Figure 4 A schematic diagram of the structure of the cross product transformation layer is shown in FIG. Figure 4 As shown, the cross product transformation layer can perform multi-layer cross product transformation. Specifically, the output vector of each layer is calculated by the following work.

[0125] x1=x0x T w0+b0+x

[0126]

[0127]

[0128] Among them, x0 is the reference complete feature vector, x is the complete feature vector of the staff member, and x l is the output vector corresponding to each layer, w0,w l are the weight vectors of the first layer and the corresponding layer, b0, b l are the first layer and the corresponding layer bias vector respectively.

[0129] After that, the last layer, that is, xl The output vector of the layer is linearly transformed to obtain the result of linear transformation. Specifically, x l The output vector of the layer is multiplied by the preset weight matrix and then added to the preset bias term to obtain x l The output vector of the layer corresponds to the result of the linear transformation.

[0130] The result of the linear transformation is then passed through the weight layer for dimensionality adjustment. Specifically, the Jinguo weight layer adjusts the result of the linear transformation to 256 dimensions to obtain the second fusion feature of the staff.

[0131] S2014. For each worker, the first fusion feature and the second fusion feature corresponding to the worker are input into the output layer for prediction processing to output the predicted completion rate of the target task corresponding to the worker.

[0132] Combined with the scenario example, the first fusion feature and the second fusion feature of each staff member are added in the Sigmoid output layer and then passed through a layer to produce a one-dimensional output. The Sigmoid activation function finally outputs the predicted completion rate of each staff member for the target task, and the predicted completion rate is in the range of 0 to 1.

[0133] S2015. Compare the predicted completion rates of the target tasks for each worker, and determine the worker with the highest predicted completion rate as the target worker.

[0134] Combined with the scenario example, the higher the predicted completion rate of a worker, the greater the probability that the worker will complete the target task, so the worker with the highest predicted completion rate is selected as the target worker.

[0135] S2016. Send the target task to the terminal corresponding to the target staff member.

[0136] Combined with the scenario example, each worker has his or her own corresponding terminal, which can be a server. The target task should be assigned to the target worker, so the target task can be sent to the terminal of the target worker so that the target worker receives the target task through the corresponding terminal.

[0137] This example predicts the completion rate of each worker for the target task, and then assigns the target task to the worker with the highest predicted completion rate. This ensures that the target task can be completed with the highest probability, thereby achieving the goal of improving the task completion rate.

[0138] Optional, Figure 5 Schematic diagram of the task allocation processing method provided in this application Figure 2 ,like Figure 5 As shown, S204 includes:

[0139] S501: Obtain a preset upper limit value and a preset lower limit value corresponding to each numerical feature information of the target customer.

[0140] Combined with the scenario example, for the target customer's age, family income, number of properties, number of vehicles, assets, liabilities, weight, height, current paid premiums and insurance coverage, etc., obtain the preset upper limit and preset lower limit corresponding to each numerical feature.

[0141] S502: Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information.

[0142] In combination with the scenario example, such as the following formula, when normalizing each numerical feature to obtain the normalized value, the absolute value of the numerical feature of the target customer is compared with the difference between the preset upper limit value and the preset lower limit value.

[0143]

[0144] For example, taking the age of the target customer as an example, if the age of the target customer is 40 years old, the preset upper limit value and the preset lower limit value of the age are 60 years old and 18 years old respectively, then the normalized value of the age of the target customer after normalization is 95%.

[0145] S503: Concatenate the ratio results corresponding to each numerical feature information of the target customer to obtain a numerical feature vector of the target customer.

[0146] Combined with the scenario example, by using the above formula, each numerical feature information of the target customer is normalized to obtain the normalized value corresponding to each numerical feature information, and the normalized values ​​corresponding to each numerical feature information are concatenated in a preset order to obtain the numerical feature vector corresponding to the target customer.

[0147] Optional, Figure 6 Schematic diagram of the task allocation processing method provided in this application Figure 3 ,like Figure 6 As shown, S208 includes:

[0148] S601, obtaining a preset upper limit value and a preset lower limit value corresponding to each numerical characteristic information of the staff member.

[0149] Combined with the scenario example, the numerical feature information of each staff member is normalized in turn in accordance with the method of processing the numerical feature information of the target customer. Specifically, for each staff member, the preset upper limit value and the preset lower limit value corresponding to each numerical feature that the staff member should correspond to are obtained.

[0150] S602: Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information.

[0151] Combined with the scenario example, the above formula is used to obtain the normalized value corresponding to each numerical feature information of the staff member.

[0152] S603: Concatenate the ratio results corresponding to each numerical feature information of the staff member to obtain a numerical feature vector of the staff member.

[0153] In combination with the scenario example, the normalized values ​​corresponding to each numerical feature information of each staff member are concatenated in a preset order to obtain the numerical feature vector corresponding to each staff member.

[0154] Based on the method provided in this example, each numerical feature information corresponding to the target customer and each staff member is normalized to avoid misleading the task completion rate prediction model due to the excessive absolute value of some numerical features.

[0155] Optional, Figure 7 Schematic diagram of the task allocation processing method provided in this application Figure 4 ,like Figure 7 As shown, in S2012, reference staff members are screened from various staff members, including:

[0156] S701. Obtain task completion records, wherein the task completion records include each completed historical task and the staff member who completed each historical task.

[0157] In combination with the scenario example, the task completion record can be in the form of a table. The historical tasks represent the sales tasks that have been assigned to the corresponding staff. The task completion record is mainly responsible for recording each completed sales task and the corresponding staff.

[0158] S702: Determine the staff member who has completed the largest number of historical tasks as the reference staff member.

[0159] In combination with the scenario example, the task completion record can be in the form of a table. For example, each month corresponds to a task completion record. The staff member with the largest number of historical tasks recorded can be determined as the reference staff member based on the completed historical tasks recorded in the task completion record of that month.

[0160] Based on the method provided in this example, a reference worker can be determined based on the task completion records.

[0161] Optionally, the embedding layer includes an autoencoder fully connected neural network model;

[0162] Accordingly, Figure 8 Schematic diagram of the task allocation processing method provided in this application Figure 5 ,like Figure 8 As shown, S206 includes:

[0163] S801, input each category feature information into the autoencoder fully connected neural network model to output the category feature vector of the target customer;

[0164] Combined with the scenario example, the embedding layer can be implemented using an autoencoder fully connected neural network model. Fig. 9 The structural diagram of the fully connected neural network model of the autoencoder is shown in Figure 2. Fig. 9 As shown, the autoencoder fully connected neural network model includes a fully connected layer of an encoder and a fully connected layer of a decoder. After each category feature information of the target customer is input into the autoencoder fully connected neural network model, it is encoded by the fully connected layer of the encoder to output a 20-dimensional encoding vector, which is used as the category feature vector of the target customer. The 20-dimensional encoding vector can pass through the fully connected layer of the decoder, and after passing through the fully connected layer of the decoder, an output vector with the same dimension as the temporary vector can be generated.

[0165] Accordingly, S209 includes:

[0166] S802. Each category feature information of the staff member is input into the autoencoder fully connected neural network model to output the category feature vector of the staff member.

[0167] Combined with the scenario example, the category feature information of the target customers is processed in the same way. For each staff member, each category feature information of the staff member is input into the fully connected neural network model of the autoencoder, and then encoded by the fully connected layer of the encoder to output a 20-dimensional encoding vector, which is used as the category feature vector of the staff member.

[0168] Based on the method provided in this example, the feature information of the target customer and each staff member can be encoded through the autoencoder fully connected neural network model to obtain the corresponding category feature vector.

[0169] Optional, Fig.10 Schematic diagram of the task allocation processing method provided in this application Figure 6 ,like Fig.10 As shown, the training process of the autoencoder fully connected neural network model includes:

[0170] S1001. Acquire first training data, where the first training data includes training category feature information.

[0171] In combination with the scenario example, the first training data is used to train the autoencoder fully connected neural network model, and the autoencoder fully connected neural network model is used to encode category feature information, so the first training data may include multiple preset training category feature information.

[0172] S1002. Input the training category feature information into the autoencoder fully connected neural network model to output an iteration result.

[0173] Combined with the scenario example, each training category feature information included in the first training data is spliced ​​in a preset order to obtain a temporary training vector. The obtained temporary training vector is input into the autoencoder fully connected neural network model, and the encoder in the autoencoder fully connected neural network model encodes the temporary training vector to obtain the first iteration result, and outputs the iteration result.

[0174] S1003. Input the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result.

[0175] Combined with the scenario example, the first iteration result is used as input and input into the autoencoder fully connected neural network model. The encoder in the autoencoder fully connected neural network model encodes the iteration result outputted for the first time to obtain a new iteration result, that is, the second iteration result, and outputs the second iteration result.

[0176] S1004, loopingly executing the process of inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result.

[0177] Combined with the scenario example, the second iteration result output by the autoencoder fully connected neural network model is used as input again and input into the autoencoder fully connected neural network model. The encoder in the encoder fully connected neural network model encodes the second iteration result and obtains a new iteration result again, that is, the third iteration result, and outputs the third iteration result. The process of taking the output iteration result as input and inputting it into the autoencoder fully connected neural network model to output a new iteration result is executed in a continuous loop.

[0178] S1005. Use a loss function to calculate the loss value between each input and the corresponding output of the autoencoder fully connected neural network model. If the loss value exceeds a first preset value, update the parameters of the autoencoder fully connected neural network model until the loss value is less than the first preset value, and then stop training the autoencoder fully connected neural network model.

[0179] Combined with the scenario example, the mean square error can be used as the loss function to calculate the loss value between each input and output during the training process of the autoencoder fully connected neural network model. The first preset value can be determined according to the actual situation. In this example, the first preset value can be determined as 2%. If the loss value between each input and output exceeds 2%, the gradient of the loss function relative to the model parameter is calculated by back propagation of the calculated loss value to update, and the loop process of S1005 is continued. If the loss value between each input and output is less than or equal to 2%, the loop process of S1005 can be stopped to obtain a trained autoencoder fully connected neural network model. Based on the method provided in this example, the training of the autoencoder fully connected neural network model can be completed.

[0180] Optional, Fig.11 Schematic diagram of the task allocation processing method provided in this application Figure 7 ,like Fig.11 As shown, the training process of the task completion rate prediction model includes:

[0181] S1101. Acquire multiple interaction records, wherein the interaction records include customer information, staff information, and interaction content between the customer and the staff.

[0182] Combined with the scenario example, the interaction record is the interaction information between the staff and the customer when completing the sales task, including the staff's personnel information and customer information corresponding to the interaction record, as well as the interaction content between the staff and the customer. The interaction content represents the communication information between the staff and the customer, which can be text information or voice information.

[0183] S1102: Determine the actual completion rate corresponding to each interaction record based on the interaction content between the customer and the staff in each interaction record.

[0184] Combined with the scenario example, the actual completion rate mentioned in this example is a preset probability and does not represent the customer's consumption of the products sold. In other words, the customer's consumption intention for the products sold can be judged based on the interaction between the customer and the staff, and the corresponding actual completion rate can be preset based on the customer's consumption intention. The actual completion rate is between 0 and 1. If the customer has a strong consumption intention in the interaction content, but the actual situation indicates that the customer has not actually bought it, then the actual completion rate is not 0, but a value predicted based on the customer's consumption intention, such as 90%.

[0185] S1103. Obtain second training data, wherein the second training data includes customer characteristic information of the customer and personnel characteristic information of the staff corresponding to each interaction record, wherein the customer characteristic information includes each numerical characteristic information and each category characteristic information of the customer, and the personnel characteristic information of the staff includes each numerical characteristic information and each category characteristic information of the staff.

[0186] In combination with the scenario example, the second training data includes multiple interaction records. According to the interaction content between the customer and the staff included in each interaction record, the actual completion rate corresponding to each interaction record is preset.

[0187] S1104: Initialize model parameters of the task completion rate prediction model.

[0188] Combined with the scenario example, before training the task completion rate prediction model, each parameter in the task completion rate prediction model is initialized to obtain the initial value of each parameter.

[0189] S1105. Input each numerical feature information and each category feature information of the customer and each numerical feature information and each category feature information of each staff member into the task completion rate prediction model to output the predicted completion rate corresponding to each staff member.

[0190] Combined with the scenario example, according to the process of the above embodiment, the processing of each interaction record in the second training data is as follows: each numerical feature information of the customer is normalized to obtain the customer's numerical feature vector, and each category feature information of the customer is encoded through the embedding layer to obtain the customer's category feature vector. Similarly, the numerical feature vector and category feature vector of the staff are obtained. After the customer's numerical feature vector and category feature vector and the staff's numerical feature vector and category feature vector are connected in the connection layer, the first fusion feature is output after processing by the first module, and the staff's numerical feature vector and category feature vector and the reference staff's numerical feature vector and category feature vector are processed by the second module to obtain the second fusion feature. The first fusion feature and the second fusion feature are processed by the Sigmoid output layer to obtain the corresponding predicted completion rate. In this way, the corresponding predicted completion rate of the staff in each interaction record in each second training data is obtained.

[0191] S1106. Calculate the loss value between the predicted completion rate and the actual completion rate corresponding to each staff member through the loss function.

[0192] Combined with the scenario example, the loss value between the predicted completion rate and the actual completion rate is calculated using the mean square error loss function.

[0193] S1107. If the loss value exceeds a second preset value, the model parameters of the task completion rate prediction model are updated until the second training data completes the training of the task completion rate prediction model.

[0194] Combined with the scenario example, the second training data can be divided into multiple batches. If there are 1000 interaction records in the second training data, the 1000 interaction records can be divided into 100 batches. When training the task completion rate prediction model, after initializing the parameters of the task completion rate prediction model, a batch is randomly selected from 100 batches as the starting point, and the task completion rate prediction model is iteratively trained. If the loss value between the predicted completion rate and the actual completion rate corresponding to the interaction record exceeds the second preset value, the parameters in the task completion rate prediction model are updated by back propagation. After the training of the task completion rate prediction model is completed for all 1000 interaction records, the training of the task completion rate prediction model is terminated to obtain a trained task completion rate prediction model. Based on the method provided in this example, the training of the task completion rate prediction model can be completed.

[0195] Fig.12 The structure diagram of the task allocation processing device provided by this application is as follows: Fig.12 As shown, the task allocation processing device includes:

[0196] An acquisition module 121 is used to acquire a target task to be assigned and determine a target customer corresponding to the target task;

[0197] The acquisition module 121 is further used to acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer;

[0198] The acquisition module 121 is further used to acquire the personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member;

[0199] The processing module 122 is used to normalize each numerical feature information of the target customer to obtain a numerical feature vector of the target customer;

[0200] The input module 123 is used to input each category feature information of the target customer into the task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model:

[0201] The input module 123 is further used to input each category feature information into the embedding layer to output the category feature vector of the target customer;

[0202] The input module 123 is further used to input the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer;

[0203] The processing module 122 is further used to normalize each numerical feature information of each staff member to obtain a numerical feature vector of the staff member;

[0204] The input module 123 is further used to input each category feature information of each staff member into the embedding layer to output the category feature vector of the staff member;

[0205] The input module 123 is further used to input the numerical feature vector and the category feature vector of each staff member into the connection layer to output a complete feature vector corresponding to the staff member;

[0206] The input module 123 is further used to input the complete feature vector corresponding to each staff member and the complete feature vector corresponding to the target customer into the first module for fusion and activation processing, so as to output the first fusion feature corresponding to the staff member;

[0207] A determination module 124, configured to select a reference worker from among the workers, and determine a complete feature vector corresponding to the reference worker as a reference complete feature vector;

[0208] The input module 123 is further used to input the complete feature vector corresponding to each staff member and the reference complete feature vector into the second module for linear transformation processing and dimensionality reduction processing, so as to output the second fusion feature corresponding to the staff member;

[0209] The input module 123 is further used to input the first fusion feature and the second fusion feature corresponding to each worker into the output layer for prediction processing, so as to output the predicted completion rate of the target task corresponding to the worker;

[0210] The determination module 124 is further used to compare the predicted completion rates of each worker for the target task, and determine the worker with the highest predicted completion rate as the target worker;

[0211] The sending module 125 is used to send the target task to the terminal corresponding to the target staff member.

[0212] Optionally, the processing module 122 is specifically configured to obtain a preset upper limit value and a preset lower limit value corresponding to each numerical feature information of the target customer;

[0213] The processing module 122 is further configured to compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value, so as to obtain a ratio result corresponding to each numerical feature information;

[0214] The processing module 122 is further configured to concatenate the ratio results corresponding to each numerical feature information of the target customer to obtain a numerical feature vector of the target customer;

[0215] The processing module 122 is further configured to normalize each numerical feature information of the staff member to obtain the numerical feature vector of the staff member, including:

[0216] The processing module 122 is further configured to obtain a preset upper limit value and a preset lower limit value corresponding to each numerical feature information of the staff member;

[0217] The processing module 122 is further configured to compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value, so as to obtain a ratio result corresponding to each numerical feature information;

[0218] The processing module 122 is further specifically used to concatenate the ratio results corresponding to each numerical feature information of the staff member to obtain the numerical feature vector of the staff member.

[0219] Optionally, the determination module 124 is specifically configured to obtain a task completion record, wherein the task completion record includes each completed historical task and a staff member who completed each historical task;

[0220] The determination module 124 is further specifically configured to determine the worker who has completed the largest number of historical tasks as the reference worker.

[0221] Optionally, the embedding layer includes an autoencoder fully connected neural network model;

[0222] Optionally, the input module 123 is specifically used to input each category feature information into the autoencoder fully connected neural network model to output the category feature vector of the target customer;

[0223] The input module 123 is specifically used to input each category feature information of the staff member into the autoencoder fully connected neural network model to output the category feature vector of the staff member.

[0224] Optionally, the processing module 122 is further configured to obtain first training data, where the first training data includes training category feature information;

[0225] The processing module 122 is further configured to input the training category feature information into the autoencoder fully connected neural network model to output an iteration result;

[0226] The processing module 122 is further configured to input the iteration result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iteration result;

[0227] The processing module 122 is further specifically used to cyclically execute the process of inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result;

[0228] The processing module 122 is specifically used to use a loss function to calculate the loss value between each input and the corresponding output of the autoencoder fully connected neural network model. If the loss value exceeds a first preset value, the parameters of the autoencoder fully connected neural network model are updated until the loss value is less than the first preset value, and the training of the autoencoder fully connected neural network model is stopped.

[0229] Optionally, the processing module 122 is further configured to obtain a plurality of interaction records, wherein the interaction records include customer information, staff information, and interaction content between the customer and the staff;

[0230] The processing module 122 is further configured to determine an actual completion rate corresponding to each interaction record based on the interaction content between the customer and the staff in each interaction record;

[0231] The processing module 122 is further specifically configured to obtain second training data, wherein the second training data includes customer feature information of a customer and staff feature information of a staff member corresponding to each interaction record, wherein the customer feature information includes each numerical feature information and each category feature information of the customer, and the staff feature information includes each numerical feature information and each category feature information of the staff member;

[0232] The processing module 122 is further specifically used to initialize the model parameters of the task completion rate prediction model;

[0233] The processing module 122 is further specifically used to input each numerical feature information and each category feature information of the customer and each numerical feature information and each category feature information of each staff member into the task completion rate prediction model to output the predicted completion rate corresponding to each staff member;

[0234] The processing module 122 is further configured to calculate the loss value between the predicted completion rate and the actual completion rate corresponding to each staff member through a loss function;

[0235] The processing module 122 is further specifically used to update the model parameters of the task completion rate prediction model if the loss value exceeds a second preset value until the second training data completes the training of the task completion rate prediction model.

[0236] The task allocation processing device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be described in detail here.

[0237] Fig.13 This is a schematic diagram of the structure of the computer device provided in this application. Fig.13As shown, the computer device 50 provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the device 50 also includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus 504.

[0238] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that at least one processor 501 executes the above method.

[0239] The specific implementation process of the processor 501 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0240] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0241] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0242] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0243] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0244] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0245] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0246] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0247] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0248] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0249] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0250] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0251] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0252] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A task allocation processing method, characterized in that: The method is applied to the server, and includes: Obtaining the target tasks to be assigned, and determining the target customers corresponding to the target tasks; Acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer; Acquire personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member; Normalizing each numerical feature information of the target customer to obtain a numerical feature vector of the target customer; Input each category feature information of the target customer into a task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model: Input each category feature information into the embedding layer to output the category feature vector of the target customer; Inputting the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer; For each staff member, normalize each numerical feature information of the staff member to obtain a numerical feature vector of the staff member; For each staff member, input each category feature information of the staff member into the embedding layer to output the category feature vector of the staff member; For each staff member, the numerical feature vector and the category feature vector of the staff member are input into the connection layer to output a complete feature vector corresponding to the staff member; For each staff member, the complete feature vector corresponding to the staff member and the complete feature vector corresponding to the target customer are input into the first module for fusion and activation processing, so as to output the first fusion feature corresponding to the staff member; Selecting a reference worker from each worker, and determining a complete feature vector corresponding to the reference worker as a reference complete feature vector; For each staff member, the complete feature vector corresponding to the staff member and the reference complete feature vector are input into the second module for linear transformation processing and dimensionality reduction processing, so as to output the second fusion feature corresponding to the staff member; For each worker, the first fusion feature and the second fusion feature corresponding to the worker are input into the output layer for prediction processing to output the predicted completion rate of the target task corresponding to the worker; Comparing the predicted completion rates of the target tasks for each worker, and determining the worker with the highest predicted completion rate as the target worker; The target task is sent to the terminal corresponding to the target worker.

2. The method according to claim 1, characterized in that The normalizing each numerical feature information of the target customer to obtain the numerical feature vector of the target customer includes: Obtaining a preset upper limit value and a preset lower limit value respectively corresponding to each numerical characteristic information of the target customer; Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information; splicing the ratio results corresponding to each numerical feature information of the target customer to obtain the numerical feature vector of the target customer; The normalizing each numerical feature information of the staff member to obtain the numerical feature vector of the staff member includes: Obtaining a preset upper limit value and a preset lower limit value corresponding to each numerical characteristic information of the staff member; Compare the absolute value of each numerical feature information with the difference between the corresponding preset upper limit value and the preset lower limit value to obtain a ratio result corresponding to each numerical feature information; The ratio results corresponding to each numerical feature information of the staff member are spliced ​​to obtain the numerical feature vector of the staff member.

3. The method according to claim 1, characterized in that The screening of reference staff from various staff members includes: Obtaining a task completion record, wherein the task completion record includes each completed historical task and the staff member who completed each historical task; The worker who has completed the largest number of historical tasks is determined as the reference worker.

4. The method according to claim 1, characterized in that: The embedding layer includes an autoencoder fully connected neural network model; Accordingly, the step of inputting each category feature information into the embedding layer to output the category feature vector of the target customer includes: Input each category feature information into the autoencoder fully connected neural network model to output the category feature vector of the target customer; Accordingly, the step of inputting each category feature information of the staff member into the embedding layer to output the category feature vector of the staff member includes: Each category feature information of the staff member is input into the autoencoder fully connected neural network model to output the category feature vector of the staff member.

5. The method according to claim 4, characterized in that The training process of the autoencoder fully connected neural network model includes: Acquire first training data, where the first training data includes training category feature information; Inputting the training category feature information into the autoencoder fully connected neural network model to output an iteration result; Inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result; Circularly executing the process of inputting the iterative result output by the autoencoder fully connected neural network model into the autoencoder fully connected neural network model to output a new iterative result; A loss function is used to calculate the loss value between each input and the corresponding output of the autoencoder fully connected neural network model. If the loss value exceeds a first preset value, the parameters of the autoencoder fully connected neural network model are updated until the loss value is less than the first preset value, and the training of the autoencoder fully connected neural network model is stopped.

6. The method according to any one of claims 1 to 5, characterized in that: The training process of the task completion rate prediction model includes: Acquire multiple interaction records, wherein the interaction records include customer information, staff information, and interaction content between the customer and the staff; Based on the interaction content between the customer and the staff in each interaction record, determine the actual completion rate corresponding to each interaction record; Acquire second training data, wherein the second training data includes customer feature information of a customer and staff feature information of a staff member corresponding to each interaction record, wherein the customer feature information includes each numerical feature information and each category feature information of the customer, and the staff feature information includes each numerical feature information and each category feature information of the staff member; Initializing model parameters of the task completion rate prediction model; Input each numerical feature information and each category feature information of the customer and each numerical feature information and each category feature information of each staff member into the task completion rate prediction model to output the predicted completion rate corresponding to each staff member; The loss function is used to calculate the loss value between the predicted completion rate and the actual completion rate of each worker. If the loss value exceeds a second preset value, the model parameters of the task completion rate prediction model are updated until the second training data completes the training of the task completion rate prediction model.

7. A task allocation processing device, characterized in that: The device comprises: An acquisition module is used to acquire the target task to be assigned and determine the target customer corresponding to the target task; The acquisition module is further used to acquire customer characteristic information of the target customer, wherein the customer characteristic information of the target customer includes each numerical characteristic information and each category characteristic information of the target customer; The acquisition module is further used to acquire the personnel characteristic information of each staff member, wherein the personnel characteristic information of the staff member includes each numerical characteristic information and each category characteristic information of the staff member; A processing module, used for normalizing each numerical feature information of the target customer to obtain a numerical feature vector of the target customer; The input module is used to input each category feature information of the target customer into the task completion rate prediction model, wherein the task completion rate prediction model includes an embedding layer, a connection layer, a first module, a second module and an output layer, wherein the first module includes a plurality of fully connected layers, and the second module includes a cross product transformation layer and a weight layer, so as to perform the following steps through the task completion rate prediction model: The input module is further used to input each category feature information into the embedding layer to output the category feature vector of the target customer; The input module is further used to input the numerical feature vector and the category feature vector of the target customer into the connection layer to output a complete feature vector corresponding to the target customer; The processing module is further used to normalize each numerical feature information of each staff member to obtain a numerical feature vector of the staff member; The input module is further used to input each category feature information of each staff member into the embedding layer to output the category feature vector of the staff member; The input module is further used to input the numerical feature vector and the category feature vector of each staff member into the connection layer to output a complete feature vector corresponding to the staff member; The input module is further used to input, for each staff member, the complete feature vector corresponding to the staff member and the complete feature vector corresponding to the target customer into the first module for fusion and activation processing, so as to output a first fusion feature corresponding to the staff member; A determination module, used for selecting a reference worker from each worker, and determining a complete feature vector corresponding to the reference worker as a reference complete feature vector; The input module is further used to input the complete feature vector corresponding to each staff member and the reference complete feature vector into the second module for linear transformation processing and dimensionality reduction processing, so as to output the second fusion feature corresponding to the staff member; The input module is further used to input the first fusion feature and the second fusion feature corresponding to each staff member into the output layer for prediction processing, so as to output the predicted completion rate of the target task corresponding to the staff member; The determination module is further used to compare the predicted completion rates of the target tasks corresponding to each worker, and determine the worker with the highest predicted completion rate as the target worker; The sending module is used to send the target task to the terminal corresponding to the target staff member.

8. A computer device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images