User value estimation method and device, electronic equipment and computer program product

Abstract

The embodiment of the invention discloses a user value estimation method and device and electronic equipment. The method comprises the steps of obtaining target user characteristics of a target user in response to a user request; in addition, the method further comprises the step of determining the target user value by a pre-estimation model based on the target user features, and the pre-estimation model is obtained by training based on the user features, the service object features and the user value. In the embodiment, in the process of estimating the target user value of the target user by using the trained estimation model according to the user request, the service object feature relevance implied in the user features learned by the model can be utilized to estimate the target user value of the target user without considering the direct influence of the service object features on the user value. The target user value is estimated more accurately according to the user features, and then targeted services are allocated according to the more accurate user value.

Description

Technical Field

[0001] The embodiments of this specification belong to the field of user value estimation, and particularly relate to a method, device, electronic device, and computer program product for estimating user value. Background Art

[0002] To improve the success rate of services like insurance, home purchases, and financial management, user value can be estimated and assigned to different service recipients based on their value, providing personalized services. Currently, user value estimation, for example, is based on pre-set service rules. These indicators are often single and highly subjective, making it difficult to accurately assess user value and thus unable to provide effective personalized services. Summary of the Invention

[0003] Embodiments of the present disclosure provide a method, apparatus, and electronic device for estimating user value, aiming to solve one or more of the above-mentioned problems and other potential problems.

[0004] According to a first aspect of the present disclosure, a method for estimating user value is provided. The method includes obtaining target user characteristics of a target user in response to a user request. The method also includes determining the target user value based on the target user characteristics using an estimation model, where the estimation model is trained based on the user characteristics, service target characteristics, and user value.

[0005] According to a second aspect of the present disclosure, a device for estimating user value is provided. The device includes a user feature acquisition module configured to obtain target user features of a target user in response to a user request. Furthermore, the device includes a user value determination module configured to determine the target user value based on the target user features using an estimation model trained based on the user features, service target features, and user value.

[0006] According to a third aspect of the present disclosure, an electronic device is provided, comprising one or more processors and a memory associated with the one or more processors, wherein the memory is used to store program instructions. When the program instructions are read and executed by the one or more processors, the method provided according to the first scheme is executed.

[0007] According to a fourth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the method provided according to the first aspect is implemented.

[0008] It should be understood that the contents described in the Summary of the Invention section are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein: Figure 1 A schematic diagram illustrating an example environment in which various embodiments of the present disclosure may be implemented; Figure 2 A schematic diagram showing a flow chart of a method for estimating user value according to some embodiments of the present disclosure; Figure 3 A schematic diagram of the architecture of the estimation model in the training phase of some embodiments of the present disclosure is shown; Figure 4 A schematic diagram showing a network structure of an estimation model according to some embodiments of the present disclosure is shown; Figure 5 A schematic diagram showing the principle process of the estimation model in the training phase of some embodiments of the present disclosure; Figure 6 Schematic diagrams showing the architecture of the estimation model in the training phase according to some further embodiments of the present disclosure are shown; Figure 7 A schematic diagram showing a network structure of an estimation model according to some embodiments of the present disclosure is shown; Figure 8 A schematic diagram showing the principle process of the estimation model in the training stage according to some other embodiments of the present disclosure is shown; Figure 9 A schematic diagram illustrating the architecture of an estimation model in the inference phase of some embodiments of the present disclosure is shown; Figure 10 A schematic diagram showing a process of obtaining user features in some embodiments of the present disclosure is provided; Figure 11 A schematic diagram showing a process flow of obtaining user characteristics according to the present disclosure; Figure 12 A schematic diagram illustrating the architecture of an estimation model in the inference phase of further embodiments of the present disclosure is shown; Figure 13 A schematic diagram showing a flow chart of a process for obtaining user features according to yet other embodiments of the present disclosure; Figure 14 A schematic diagram showing a process flow of an estimation model selection process in some embodiments of the present disclosure is shown; Figure 15 A schematic diagram showing a process of providing services to target users according to their value in some embodiments of the present disclosure is provided; Figure 16 A schematic diagram showing the structure of a device for estimating user value according to some embodiments of the present disclosure is shown; Figure 17A schematic block diagram of an electronic device according to some embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0010] To make the purpose, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0011] The terms "including" and "having" and any variations thereof in this specification and claims and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or apparatuses. Depending on the context, the word "if" as used herein may be interpreted as "at..." or "when..." or "in response to determining" or "in response to detecting".

[0012] It should be noted that the information, data, and instructions involved in this application are authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0013] As mentioned above, in order to improve the success rate of services, it is necessary to assign appropriate service recipients to users based on their value in order to provide personalized services to them. Currently, user value can be estimated based on, for example, preset service rules. This involves inputting user data based on pre-set service indicators and their evaluation rules, scoring various service indicators, and then determining user value based on the scoring results. However, this approach often uses a single indicator and is highly subjective. Furthermore, many factors influence user value, and the resulting user value is often inaccurate. This results in inappropriate service recipients being assigned to users, making it impossible to provide effective personalized services to users.

[0014] To address this issue, the embodiments of the present disclosure propose a user value estimation solution. In this embodiment, the impact of user characteristics and service target characteristics can be considered in advance to train an estimation model. Then, after the target user for whom user value estimation is required is determined based on the user request, the model prediction process does not consider service target characteristics, but instead considers the target user's target user characteristics to determine the target user value.

[0015] Through the above approach, the estimation model can learn the complex correlations between user characteristics and service object characteristics during the training phase, capturing possible implicit relationships between data, improving the model's generalization and robustness. Furthermore, when using the trained estimation model to estimate the target user's value based on user requests, the model leverages the implicit correlations with service object characteristics within the user characteristics learned, allowing for a more accurate estimation of the target user's value based on the user characteristics. In this way, the user value of each type of user is estimated primarily based on the user's personalized characteristics, without significant deviations due to differences in service objects. Furthermore, because the estimation model learns the correlations between users, service objects, and user value, it can consider the indirect impact of service objects when no specific service objects are assigned. This allows for more accurate estimation results for all types of users under equal influence, enabling more accurate personalized services to be provided to users, improving both user experience and service success rates.

[0016] Figure 1 1 shows a schematic diagram of an example environment 100 in which various embodiments of the present disclosure may be implemented. Figure 1As shown, environment 100 may include target user 110, terminal 120, and server 130. Terminal 120 may be an electronic device with communication capabilities, including but not limited to smartphone 120a, tablet computer 120b, in-vehicle device, handheld device, wearable device, personal computer 120n, etc. Server 130 may be a single server device, such as a rack-mounted, blade, tower, or cabinet-mounted server device, or a hardware device with strong computing power, such as a workstation or mainframe computer. It may also be a server cluster composed of multiple servers. Each server in the server cluster may be symmetrically arranged, with each server having equivalent functionality and status in the transaction chain. Each server can independently provide external services, and independent services can be understood as providing services without the assistance of other servers. While using terminal 120, user 110 can actively or passively obtain information about services that may be of interest, such as insurance, home purchase, and financial management, through access, query, and search, and in this process, feedback user requests 131 to server 130. Based on the user request 131, the server 130 can determine the source of the request, the requested service, and other information, and call the target user features 132 of the user 110 through the corresponding source or interface. By inputting the target user features 132 into the estimation model 133 stored in the server 130 and pre-trained based on the user features, service object features, and user value, the estimation model 133 outputs the target user value 134. The estimation model 133 is pre-trained based on the user features, service object features, and user value, and can be trained on the server 130 or deployed on the server 130 after being trained by other servers. The target user value 134 can more accurately represent the objective value of the user before a service object is assigned to intervene, so that corresponding personalized services can be provided to the user 110 based on the target user value 134.

[0017] It should be noted that the terminal 120 and the server 130 communicate interactively via a network connection. The network can be a wireless network or a wired network. Wireless networks include, but are not limited to, cellular networks, wireless local area networks, infrared networks, or Bluetooth networks. Wired networks include, but are not limited to, Ethernet, universal serial bus (USB), or controller area network. In one or more embodiments of this specification, technologies and / or formats such as Hypertext Markup Language (HTML) and Extensible Markup Language (XML) are used to represent data exchanged over the network. Conventional encryption technologies such as Secure Sockets Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used in place of or in addition to the aforementioned data communication technologies.

[0018] Figure 2 FIG2 shows a flow chart of a method 200 for estimating user value according to some embodiments of the present disclosure. The method 200 may be performed by, for example Figure 1 The server 130 in the embodiment executes. Figure 2 As shown, at block 202, method 200 can respond to a user request and obtain target user characteristics of a target user. In this embodiment, when a target user accesses any service through a terminal, the terminal can generate a user request indicating that the target user may be interested in the service and send the user request to the server. The terminal can access the service by, for example, accessing a link corresponding to the service, accessing an advertising page corresponding to the service, initiating a consultation session corresponding to the service, searching for keywords related to the service in a designated application, or launching an application related to the service. Depending on the method of accessing the service, the user request can be actively or passively generated by the target user. The target user can be a new user accessing the service for the first time or a retained user who has used the service before. After receiving the user request, the server will respond to the user request and obtain the target user characteristics of the target user from the corresponding database of the service corresponding to the user request. As an example, user characteristics can include attributes such as the user's ID and occupation, as well as behavioral characteristics such as the number of historical visits to the service, the number of historical purchases, and the historical purchase amount.

[0019] In block 204, method 200 further includes determining the target user value using an estimation model based on target user characteristics. The estimation model is trained based on user characteristics, service target characteristics, and user value. In this embodiment, the server may store a service-specific estimation model that has been pre-trained using user characteristics and service target characteristics as inputs and user value as output. Service target characteristics may include attributes such as the service target's ID (i.e., the person actually providing service to the user), job level, and service preferences (e.g., preferred service type or preferred user type). They may also include the short-term response rate (e.g., response rate within 30 seconds), service follow-up rate (e.g., frequency of at least two follow-up services within three days), number of users historically served, and number of historical orders. By inputting only the target user characteristics into the estimation model, the target user value can be determined without assigning a specific service target and excluding the influence of the service target.

[0020] Through the above method, the estimation model takes into account user characteristics and service object characteristics during training, allowing the model to learn the complex correlations between user characteristics and service object characteristics and more comprehensively understand the data distribution patterns. However, in the actual reasoning stage, since no specific service object has been assigned at this time, the influence of the service object needs to be excluded, and user value will be predicted only based on user characteristics. This embodiment can ignore the influence of service object characteristics during the reasoning stage, while implicitly utilizing the distribution characteristics of user characteristics learned by the model during training to indirectly reflect the correlation of service object characteristics, thereby improving the accuracy of user value prediction and achieving more efficient and accurate prediction.

[0021] Figure 3 The following is a schematic diagram of the architecture of the estimation model 300 during the training phase according to some embodiments of the present disclosure. In the estimation model 300, the model includes a user feature extraction module 311, a service object feature extraction module 321, and a prediction module 330. The user feature extraction model 311 can, for example, extract local features through a convolutional layer and reduce dimensionality through a pooling layer to convert user features 310 into a user feature vector 312, and transmit the user feature vector 312 to the prediction module 330. Similarly, the service object feature extraction module 321 can convert service object features into a service object feature vector 322, and transmit the service object feature vector 322 to the prediction module 330. The prediction module 330 can perform a nonlinear transformation on the feature vector using an activation function (such as ReLU, Sigmoid, etc.) through weighted summation to output a user value 340.

[0022] Figure 4A schematic diagram of the network structure of the estimation model 400 of some embodiments of the present disclosure is shown. In the estimation model 400, at least an embedding layer, a splicing layer, a hidden layer, and an output layer (not shown in the figure) may be included. The embedding layer is used to convert high-dimensional sparse features (such as user features 410 or service object features 420) into low-dimensional dense vectors, thereby retaining semantic information and reducing computational complexity. The splicing layer is used to expand feature vectors from different sources in dimension (such as channel splicing) to retain information of all features. The hidden layer is used to perform nonlinear transformation and feature extraction on the input features to capture more complex patterns. The output layer is used to output user value 430.

[0023] Figure 5 A schematic diagram illustrates the principle process 500 of the training phase of an estimation model in some embodiments of the present disclosure. Historical service data can store user information recorded during each service (regardless of the ultimate success of the service), information about the service recipients who provided the service to the user, and the results of the service. User feature samples 510-1 and service recipient feature samples 510-2 can be directly obtained using the user information and service recipient information stored in the historical service data. Furthermore, based on information such as service difficulty, service duration, and service results in each historical service data set, user value can be manually annotated for each historical service data set to obtain user value samples 510-3. Training samples 510 can be constructed based on user feature samples 510-1, service recipient feature samples 510-2, and user value samples 510-3. The initial estimation model used in training can be a convolutional neural network model, a recurrent neural network model, or the like. During the training of the estimation model 520 using the training samples 510, a model generator 521 can generate predicted user value 522 based on the user feature samples 510-1 and service recipient feature samples 510-2. Based on the comparison between the user value sample 510-3 and the predicted user value 522, the generator loss 523 can be obtained, and the generator loss 523 is used to make a loss judgment on the predicted user value 522. The loss judgment can be judged using the comparison loss function. The generator loss 523 can be a large value, and then the generator loss 523 can be back-propagated to the generator 521 to guide the optimization of the parameters of the generator 521, thereby realizing a round of supervised training of the generator 521. Such a training process can be iteratively executed round after round until the generator 521 can generate a more accurate predicted user value 522, that is, the loss value calculated by the loss function is smaller. After the training is completed, the estimation model 520 can output the user value 530.

[0024] Figure 6The following diagram illustrates the architecture of an estimation model 600 during the training phase, according to yet other embodiments of the present disclosure. The estimation model 600 includes a user feature extraction module 611, a service object feature extraction module 621, a service-related feature extraction module 631, and a prediction module 640. The user feature extraction module 611 can, for example, convert user features 610 into user feature vectors 612 by using convolutional layers to extract local features and pooling layers to reduce dimensionality, and transmit the user feature vectors 612 to the prediction module 640. Similarly, the service object feature extraction module 621 and the service-related feature extraction module 631 can convert service object features and service-related features into service object feature vectors 622 and service-related feature vectors 632, respectively, and transmit these service object feature vectors 622 and service-related feature vectors 632 to the prediction module 640. The prediction module 640 can perform a nonlinear transformation on the feature vectors using a weighted summation and an activation function (such as ReLU or Sigmoid) to output a user value 650.

[0025] Figure 7 A schematic diagram of the network structure of the estimation model 700 of some embodiments of the present disclosure is shown. In the estimation model 700, at least an embedding layer, a splicing layer, a hidden layer, and an output layer (not shown in the figure) may be included. The embedding layer is used to convert high-dimensional sparse features (such as user features 710, service object features 720, or service-related features 730) into low-dimensional dense vectors, thereby retaining semantic information and reducing computational complexity. The splicing layer is used to expand feature vectors from different sources in dimension (such as channel splicing) to retain information of all features. The hidden layer is used to perform nonlinear transformation and feature extraction on the input features to capture more complex patterns. The output layer is used to output user value 740.

[0026] Figure 8A schematic diagram illustrates the principle process 800 of the training phase of an estimation model according to yet other embodiments of the present disclosure. In process 800, historical service data may store user information recorded during each service (regardless of the ultimate success of the service), information about the service object provided to the user, service-related information during the service process, and the results of the service. User feature samples 810-1, service object feature samples 810-2, and service-related feature samples 810-3 may be directly obtained using the user information, service object information, and service-related information stored in the historical service data. Furthermore, based on information such as service difficulty, service duration, and service results in each historical service data set, corresponding user value may be manually annotated for each historical service data set to obtain user value samples 810-4. Training samples 810 may be constructed based on user feature samples 810-1, service object feature samples 810-2, service-related feature samples 810-3, and user value samples 810-4. The initial estimation model used in training may be a convolutional neural network model, a recurrent neural network model, or the like. During the training process of the estimation model 820 using training samples 810, the model's generator 821 can generate a predicted user value 822 based on user feature samples 810-1, service object feature samples 810-2, and service-related feature samples 810-3. By comparing user value samples 810-4 with the predicted user value 822, a generator loss 823 can be obtained. This generator loss 823 is used to determine the loss of the predicted user value 822, which can be determined using a comparison loss function. Generator loss 823 can be a large value, which can then be backpropagated to generator 821 to guide the optimization of generator 821 parameters, completing a round of supervised training for generator 821. This training process can be iteratively executed until generator 821 is able to generate more accurate predicted user values ​​822, i.e., until the loss value calculated by the loss function is smaller. After training is completed, the estimation model 820 can output user value 830.

[0027] Figure 9A schematic diagram of the architecture of the estimation model of the reasoning stage of some embodiments of the present disclosure is shown. In the estimation model, the method of not considering the influence of the service object features in the reasoning stage can be to input the target user features 910 while inputting the service object features 920 that are set to zero or empty, or to input the target user features 310 while generating the model control instructions 330. The model control instructions 330 can control the estimation model 940-1 to reset the weights corresponding to the service object feature extraction module to zero or empty during the reasoning stage. In addition, the service object extraction module can be directly removed in the definition stage after the estimation model is trained, or the service object extraction module can be skipped through code logic during reasoning to achieve the removal of the service object extraction module in the reasoning stage. The estimation model 940-2 after removal only needs to input the target user features 910. After the above processing, the estimation model can output the target user value 950.

[0028] Figure 10 A flowchart illustrating a process 1000 for obtaining user characteristics in accordance with some embodiments of the present disclosure is shown. In process 1000, a user accesses a session, link, or other information associated with a specific service, triggering a user request 1010 corresponding to that service. By parsing user request 1010, the server can determine the request object 1020 that initiated the request. Request object 1020 can, for example, be the specific service product requested (e.g., insurance, home purchase, financial services, etc.). Next, the server can query the target user's user history profile 1021 associated with the request object. User history profile 1021 refers to a user feature model constructed based on the user's past behavior, preferences, attributes, and other data, describing the user's long-term interests and characteristics. For example, user history profile 1021 can be obtained by classifying tags in an application and constructing a user-tag matrix based on the user's interest in each tag. It can also be obtained by analyzing the user's data in the application using an RFM model, or by performing cluster analysis on the user's data in the application.

[0029] In addition, the user historical profile can be the user historical profile obtained from all applications associated with the service, and then the user historical profile with the highest average similarity to other user historical profiles is retained from each user historical profile as the user historical profile to be used. For example, for a specific service A, this service A may be accessible through different applications (for example, there are application A, application B, and application C, and multiple applications can be accessed and obtain user historical profiles with user authorization). If the user accesses service A through application A, in addition to obtaining the user historical profile from application A, the server can also obtain the user historical profiles of application B and application C at the same time, and then retain the user historical profile with the highest average similarity to other user historical profiles from each user historical profile as the user historical profile to be used. The user historical profile can also be the user historical profile obtained from the application with the highest degree of association with the service.

[0030] From the user history profile 1021, data categorized as basic information 1022 and data categorized as behavioral information 1023 can be obtained. The data in basic information 1022 will be used as user attribute features, and the data in behavioral information 1023 will be used as user behavioral features to obtain target user features 1030 for the target user. User attribute features may include the user's ID, occupation, etc., while user behavioral features may include the user's historical number of visits to the service, historical number of purchases, and historical purchase amounts.

[0031] Figure 11 A flowchart illustrating a user value estimation process 1100 for some embodiments of the present disclosure is provided. In process 1100, while responding to a user request and obtaining target user features 1110 of the target user, service-related features 1111 may also be obtained. Service-related features 1111 are features of external influencing factors that are associated with the service. Service-related features 1111 may, for example, include channel features that characterize the source triggering the user value estimation, time features that characterize the current date and time, including whether it is a holiday, and weather features that characterize the current weather. During the training phase, the estimation model 1120 is trained using user features, service object features, service-related features, and user value. During the inference phase, after eliminating the influence of service object features, the estimation model 1120 inputs both the target user features 1110 and the service-related features 1111. This further improves the accuracy of the output of the estimation model 1120, ultimately resulting in the target user value 1130.

[0032] Figure 12A schematic diagram of the architecture of the estimation model for the inference phase of some other embodiments of the present disclosure is shown. In the estimation model, the influence of service object features can be ignored during the inference phase by inputting the target user features 1210 and the service-related features 1230 while inputting the service object features 1220 that are set to zero or empty. Alternatively, the target user features 1210 and the service-related features 1230 can be input while generating a model control instruction 1240. The model control instruction 1240 can control the estimation model 1250-1 to reset the weights corresponding to the service object feature extraction module to zero or empty during the inference phase. In addition, the service object extraction module can be directly removed during the definition phase after the estimation model is trained, or the service object extraction module can be skipped during inference through code logic to achieve the removal of the service object extraction module during the inference phase. After the removal, the estimation model 1250-2 only needs to input the target user features 1210 and the service-related features 1230. After performing the above processing, the estimation model can output the target user value 1260.

[0033] Figure 13 A flowchart illustrating a process 1300 for obtaining user features according to yet other embodiments of the present disclosure is shown. In process 1300, after receiving a user request 1310, the server may obtain service-related features. Different types of service-related features may require different acquisition methods, so each type of service-related feature may be acquired separately. As an example, the user request 1310 may be parsed through methods such as HTTP header verification and API key authentication to determine which client sent the user request 1310. This client will serve as the request source 1320, and the channel features 1321 of the request source 1320 will be obtained. Channel features 1321 may include the channel's request path, channel type (e.g., mobile application, webpage, etc.), channel name, channel ID, etc. Subsequently, a system access instruction 1330 may be generated, which is used to access the server system to obtain time information from the system clock and use it as time features 1331. In addition, a weather port calling instruction 1340 for a client capable of disclosing weather information may be generated, so as to access the port of the client according to the weather port calling instruction 1340 and call the corresponding weather information from the client as a weather feature 1341 .

[0034] Figure 14A flow chart of the estimation model selection process 1400 of some embodiments of the present disclosure is shown. In the process 1400, different types of services can be distinguished according to the service product category. For different types of services, a separate estimation model can be trained for each service, and each estimation model can be stored in the database 1430, and a mapping relationship between each service product category and the estimation model can be constructed. By parsing the user request 1410, the server can determine what kind of service this request corresponds to, that is, determine the service product category 1420 of this time. In the database 1430, the service product category 1420 is queried to obtain the estimation model 1440 corresponding to the service product category 1420.

[0035] As an example, assume that insurance service products correspond to service product category a, home purchase service products correspond to service product category b, and financial management service products correspond to service product category c. Prediction model a can be trained separately for insurance service products, prediction model b can be trained separately for home purchase service products, and prediction model c can be trained separately for financial management service products. The prediction models corresponding to different service products use different training samples during training, and the degree to which user characteristics and service object characteristics affect user value may also be different. The trained models will be stored in the database 1430. After the specific service product category 1420 is determined according to the user request 1410, the prediction model 1440 that matches the service product category 1420 can be queried in the database 1430.

[0036] Figure 15A flowchart illustrating a process 1500 for providing services to a target user based on their target user value, according to some embodiments of the present disclosure, is shown. In process 1500, users can be categorized based on their user value to rationally allocate service resources and provide personalized services to different types of users. Specifically, after determining the target user value 1510, in block 1520, a preset value threshold can be used to compare the target user value. The user value and the value threshold can be represented by a score derived from a scoring system or by an estimated potential transaction amount. If the target user value is not less than the value threshold, the target user's service value is considered high, indicating that the target user's situation is highly compatible with the service content, and they may have a high demand for the service, resulting in a high probability of a final transaction. Therefore, a service object set 1540 corresponding to the service product category can be queried on the server, and a service object 1541 can be assigned to the target user from the service object set 1540. This allows the service object 1541 to directly communicate with the target user via text message, phone call, or conversation, for specific communication regarding the service content. As an example, the method for allocating service objects 1541 can be to randomly select service objects 1541 from the service object set 1540 for allocation. Alternatively, the method can be to determine target service objects in the service object set 1540 that are willing to serve the target user based on the service preferences of each service object, and then randomly select from the target service objects. Alternatively, specific service objects can be allocated based on the number of users currently being served, so that the target service object currently serving the fewest users is given priority. Alternatively, the target user's user information can be sent and pushed to all service objects in the service object set 1540, and the service objects can choose whether to serve the target user on a first-come, first-served basis. Other possible implementation methods are also possible. In addition, if the target user's value is lower than the value threshold, it indicates that the target user's situation has a low degree of match with the service content of the service. The target user is likely not interested in the service, or may have simply accessed the content related to the service by accident. In this case, no service object can be allocated to the target user. Instead, service promotion information 1530 related to the service can be generated. The service promotion information 1530 can include information such as the content description and price of the service. The service promotion information 1530 will be pushed to the terminal used by the target user 1550. If the target user 1550 generates feedback information indicating interest in the service through an operation in the promotion interface of the service promotion information 1530 after receiving the service promotion information 1530, the server can respond to the feedback information and assign a service object to the target user 1550.

[0037] Figure 16The following is a schematic diagram showing the structure of a user value estimation device 1600 of some embodiments of the present disclosure. Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. Figure 16 As shown, apparatus 1600 includes a user feature acquisition module 1601 configured to respond to a user request and acquire target user features of a target user. Furthermore, the apparatus includes a user value determination module 1602 configured to determine the target user value based on the target user features using an estimation model trained based on the user features, service target features, and user value.

[0038] The user value determination module 1602 is further configured to determine the target user value based on the target user characteristics and the service object characteristics that are set to zero or empty by an estimation model; or, based on the target user characteristics, determine the target user value by the estimation model without the service object characteristic extraction module; or, based on the target user characteristics, determine the target user value by the estimation model with the corresponding weights of the service object characteristic extraction module reset to zero or empty.

[0039] User characteristics include user attribute characteristics and user behavior characteristics. User characteristic acquisition module 1601 is further configured to respond to user requests, determine the target user's user history profile under the requested object; obtain the target user attribute characteristics of the target user based on the basic information in the user history profile; and obtain the target user behavior characteristics of the target user based on the behavior information in the user history profile.

[0040] The user feature acquisition module 1601 is further configured to respond to user requests and acquire service-related features of the target user. Service-related features are features of external influencing factors that are associated with the service. The user value determination module 1602 is further configured to determine the target user's value based on the target user features and service-related features using an estimation model trained based on user features, service object features, service-related features, and user value.

[0041] The user value determination module 1602 is further configured to determine the target user value by an estimation model based on the target user characteristics, the service object characteristics that are set to zero or empty, and the service-related characteristics; or, based on the target user characteristics and the service-related characteristics, determine the target user value by an estimation model without the service object characteristic extraction module; or, based on the target user characteristics and the service-related characteristics, determine the target user value by an estimation model with the corresponding weights of the service object characteristic extraction module reset to zero or empty.

[0042] Service-related features include channel features, time features, and weather features. The user feature acquisition module 1601 is further configured to respond to user requests and acquire channel features based on the request source; acquire time features based on system access instructions; and acquire weather features based on weather port call instructions.

[0043] The device also includes a product type determination module configured to respond to a user request and determine the category of the service product currently requested. The device also includes an estimation model selection module configured to select an estimation model based on the service product category.

[0044] The device also includes a first judgment module configured to allocate a service object to the target user from a set of service objects corresponding to the service product category in response to the target user's value being not less than a preset value threshold.

[0045] The device also includes a second judgment module configured to obtain service promotion information corresponding to the service product category in response to the target user's value being lower than a preset value threshold, so as to push the service promotion information to the target user.

[0046] The estimation model includes a user feature extraction module for extracting user features, a service object feature extraction module for extracting service object features, and a prediction module for outputting user value according to the user features and the service object features.

[0047] The apparatus further includes a first sample determination module configured to determine training samples based on historical service data, the training samples including user feature samples, service object feature samples, and user value samples corresponding to the feature samples. The apparatus further includes a first prediction module configured to generate a predicted user value from an initial estimation model based on the user feature samples and service object feature samples. The apparatus further includes a first training module configured to perform at least one round of model training on the initial estimation model using the user value samples as a supervisory signal to obtain an estimation model.

[0048] The first sample determination module is also configured to obtain user feature samples and service object feature samples based on historical service data; associate user feature samples and service object feature samples corresponding to the same service into data pairs, set user value samples for the data pairs; and construct a training set based on the data pairs and user value samples.

[0049] The estimation model includes a user feature extraction module for extracting user features, a service object feature extraction module for extracting service object features, a service association feature extraction module for extracting service association features, and a prediction module for outputting user value based on user features, service object features and service association features.

[0050] The device also includes a second sample determination module configured to determine training samples based on historical service data, the training samples including user feature samples, service object feature samples, service-related feature samples, and user value samples corresponding to the feature samples. The device also includes a second prediction module configured to generate a predicted user value from an initial estimation model based on the user feature samples, service object feature samples, and service-related feature samples. The device also includes a second training module configured to perform at least one round of model training on the initial estimation model using the user value samples as a supervisory signal to obtain an estimation model.

[0051] The second sample determination module is also configured to obtain user feature samples, service object feature samples and service-related feature samples based on historical service data; associate the user feature samples, service object feature samples and service-related feature samples corresponding to the same service into data pairs, set user value samples for the data pairs; and construct a training set based on the data pairs and user value samples.

[0052] In the above embodiments, all or part of the embodiments can be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state drive (SSD)).

[0053] Figure 17 1 shows a block diagram of an electronic device 1700 that can implement various embodiments of the present disclosure. Figure 17As shown, electronic device 1700 includes a processor 1710, a disk drive 1720, an input / output interface 1730, a network interface 1740, and a memory 1750. The processor 1710, disk drive 1720, input / output interface 1730, network interface 1740, and memory 1750 can be communicatively connected via a communication bus 1760.

[0054] Among them, the processor 1710 can be implemented by a general-purpose CPU, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., to execute relevant programs to implement the technical solutions provided in this application.

[0055] The memory 1750 can be implemented in the form of ROM (Read Only Memory), RAM (Read Access Memory), static memory, dynamic storage devices, etc. The memory 1750 can store an operating system 1751 for controlling the operation of the electronic device 1700 and a basic input and output system (BIOS) 1752 for controlling the low-level operations of the electronic device 1700. In addition, a web browser 1753, a data storage management system 1754, etc. can also be stored. In short, when the technical solutions provided in this application are implemented through software or firmware, the relevant program code is stored in the memory 1750 and is called and executed by the processor 1710.

[0056] Input / output interface 1730 is used to connect to an input / output module to enable information input and output. The input / output module can be configured as a component within the device (not shown) or can be externally connected to the device to provide corresponding functions. Input devices may include a keyboard, mouse, touch screen, microphone, various sensors, etc. Output devices may include a display, speaker, vibrator, indicator light, etc.

[0057] Network interface 1740 is used to connect to a communication module (not shown) to enable communication between the device and other devices. The communication module can communicate via wired means (e.g., USB, network cable, etc.) or wireless means (e.g., mobile network, Wi-Fi, Bluetooth, etc.).

[0058] The bus 1760 comprises a pathway for transmitting information between the various components of the device (eg, the processor 1710 , the disk drive 1720 , the input / output interface 1730 , the network interface 1740 , and the memory 1750 ).

[0059] It should be noted that although the above device only shows the processor 1710, disk drive 1720, input / output interface 1730, network interface 1740, memory 1750, bus 1760, etc., in a specific implementation, the device may also include other components necessary for normal operation. In addition, those skilled in the art will understand that the above device may only include the components necessary to implement the method of the present application, and does not necessarily include all the components shown in the figure.

[0060] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0061] In the context of this disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. Furthermore, although operations are depicted in a particular order, this should be understood as requiring that such operations be performed in the particular order shown or in a sequential order, or that all illustrated operations be performed to achieve the desired result. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of separate embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented in multiple implementations individually or in any suitable subcombination.

[0062] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. A method for estimating user value (200), the method comprising: Respond to user requests and obtain target user characteristics of target users; as well as Based on the target user characteristics, the target user value is determined by an estimation model, and the estimation model is trained based on user characteristics, service object characteristics and user value.

2. The method (200) according to claim 1, wherein determining the target user value by an estimation model based on the target user characteristics comprises: Determining the target user value by an estimation model based on the target user characteristics and the service object characteristics that are set to zero or blank; or, Based on the target user characteristics, the target user value is determined by removing the estimation model of the service object feature extraction module; or Based on the target user characteristics, the target user value is determined by an estimation model in which the weights corresponding to the service object feature extraction module are reset to zero or empty.

3. The method (200) according to claim 1, wherein the user characteristics include user attribute characteristics and user behavior characteristics; The step of responding to the user request and obtaining target user characteristics of the target user includes: Respond to user requests and determine the target user's user history profile under the request object; Based on the basic information in the user history portrait, obtaining target user attribute characteristics of the target user; and Based on the behavior information in the user history portrait, the target user behavior characteristics of the target user are obtained.

4. The method (200) according to claim 1, further comprising: In response to a user request, obtaining a service-related feature of the target user, wherein the service-related feature is a feature of an external influencing factor that is associated with the service; as well as Based on the target user characteristics and the service-related characteristics, the target user value is determined by the estimation model, and the estimation model is trained based on the user characteristics, the service object characteristics, the service-related characteristics and the user value.

5. The method (200) according to claim 4, wherein determining the target user value by the estimation model based on the target user characteristics and the service-related characteristics comprises: Determining the target user value by an estimation model based on the target user characteristics, the service object characteristics that are set to zero or empty, and the service-related characteristics; or, Based on the target user characteristics and the service-related characteristics, the target user value is determined by removing the estimation model of the service object feature extraction module; or Based on the target user characteristics and the service-related characteristics, the target user value is determined by an estimation model in which the weights corresponding to the service object feature extraction module are reset to zero or empty.

6. The method (200) according to claim 4, wherein the service-related characteristics include channel characteristics, time characteristics and weather characteristics; The step of responding to the user request and obtaining the service-related characteristics of the target user includes: Responding to a user request, and obtaining the channel characteristics based on the request source; Based on the system access instruction, obtaining the time feature; and The weather characteristics are obtained based on the weather port call instruction.

7. The method (200) according to any one of claims 1 to 6, further comprising: In response to the user request, determine the service product category currently requested; as well as Based on the service product category, an estimation model is selected.

8. The method (200) according to claim 7, further comprising: In response to the target user value being not lower than a preset value threshold, a service object is allocated to the target user from the service object set corresponding to the service product category.

9. The method (200) according to claim 7, further comprising: In response to the target user value being lower than a preset value threshold, service promotion information corresponding to the service product category is obtained to push the service promotion information to the target user.

10. The method (200) according to any one of claims 1-3, wherein the estimation model comprises a user feature extraction module for extracting user features, a service object feature extraction module for extracting service object features, and a prediction module for outputting user value based on the user features and the service object features.

11. The method (200) according to claim 10, further comprising: Determine training samples based on historical service data, wherein the training samples include user feature samples and service object feature samples, as well as user value samples corresponding to the feature samples; Based on the user feature sample and the service object feature sample, generating a predicted user value by an initial estimation model; and The user value sample is used as a supervisory signal to perform at least one round of model training on the initial estimation model to obtain an estimation model.

12. The method (200) according to claim 11, wherein determining the training samples based on the historical service data comprises: Based on historical service data, obtain user feature samples and service object feature samples; Associating the user feature sample and the service object feature sample corresponding to the same service into a data pair, and setting a user value sample for the data pair; and A training set is constructed based on the data pairs and the user value samples.

13. According to the method (200) described in any one of claims 4 to 6, the estimation model comprises a user feature extraction module for extracting user features, a service object feature extraction module for extracting service object features, a service association feature extraction module for extracting service association features, and a prediction module for outputting user value based on the user features, the service object features, and the service association features.

14. The method (200) according to claim 13, further comprising: Determine training samples based on historical service data, wherein the training samples include user feature samples, service object feature samples, service-related feature samples, and user value samples corresponding to the feature samples; Based on the user feature sample, the service object feature sample and the service-related feature sample, an initial estimation model generates a predicted user value; and The user value sample is used as a supervisory signal to perform at least one round of model training on the initial estimation model to obtain an estimation model.

15. The method (200) according to claim 14, wherein determining the training samples based on historical service data comprises: Based on historical service data, obtain user feature samples, service object feature samples, and service-related feature samples; Associating the user feature sample, the service object feature sample, and the service-related feature sample corresponding to the same service into a data pair, and setting a user value sample for the data pair; and A training set is constructed based on the data pairs and the user value samples.

16. A user value estimation device (1600), comprising: A user feature acquisition module (1601) is configured to respond to a user request and acquire target user features of a target user; as well as The user value determination module (1602) is configured to determine the target user value based on the target user characteristics by an estimation model, wherein the estimation model is trained based on the user characteristics, the service object characteristics and the user value.

17. The apparatus (1600) according to claim 16, wherein the user feature acquisition module (1601) is further configured to: In response to a user request, obtaining a service-related feature of the target user, wherein the service-related feature is a feature of an external influencing factor that is associated with the service; The user value determination module is further configured to: Based on the target user characteristics and the service-related characteristics, the target user value is determined by the estimation model, and the estimation model is trained based on the user characteristics, the service object characteristics, the service-related characteristics and the user value.

18. An electronic device comprising: one or more processors, and A memory associated with the one or more processors, the memory being used to store program instructions, wherein when the program instructions are read and executed by the one or more processors, the steps of the method according to any one of claims 1 to 15 are executed.

19. Computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 15.

Images