Product recommendation method and device, server and medium
By combining static basic information, dynamic behavior information and risk information, the problem of insufficient accuracy of insurance product recommendations in the prior art is solved, and more accurate user demand matching and risk control are achieved.
Patent Information
- Application Number
- CN202510007615.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-03
- Publication Date
- 2025-04-08
AI Technical Summary
In the prior art, the accuracy of insurance product recommendations is insufficient, and it is difficult to fully reflect the current needs of consumers.
The multi-model fusion method is adopted to process the static basic information of the target user through the first recommendation model, and the second recommendation model processes dynamic behavior information, and combines the risk information of the risk prediction model to conduct comprehensive screening to improve the accuracy of recommendation.
It improves the accuracy of insurance product recommendations, can more accurately match the needs of target users, and reduces the risks of the recommendation system.
Smart Images
Figure CN119784475A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of big data analysis, and in particular to a product recommendation method, device, server and medium. Background Art
[0002] With the advent of the Internet era, the online trend of the insurance market has become increasingly obvious. Especially in the Internet era, consumers are more inclined to obtain and use insurance services through online channels. In today's information explosion, consumers tend to pursue clear and concise insurance product information, and hope to obtain customized insurance products and services that meet personal needs with the help of intelligent recommendation algorithms and data analysis technology.
[0003] The recommendation logic in the related art is: recommendations are made based on consumers' personal information and historical purchase behavior. However, since insurance products are not high-frequency consumer products, consumers' historical purchase behavior can only reflect consumers' previous needs to a certain extent, but cannot fully reflect consumers' current needs. This makes the accuracy of the recommended products determined by the related art for consumers insufficient, making it difficult for consumers to match products that meet their actual needs.
[0004] Therefore, how to improve the accuracy of recommended products is an urgent problem that needs to be solved. Summary of the invention
[0005] Based on this, it is necessary to provide a product recommendation method, device, server and medium that can improve the accuracy of recommended products in response to the above technical problems.
[0006] In a first aspect, the present application provides a product recommendation method, comprising:
[0007] Using the first recommendation model to process the static basic information of the target user, obtaining a first recommendation list including a plurality of first recommendation tags;
[0008] Processing the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0009] Processing the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label;
[0010] Filtering the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag;
[0011] Recommend products to the target user based on the target recommendation list.
[0012] In one embodiment, the training process of the first recommendation model includes:
[0013] Obtain static basic information and historical purchase information of multiple historical users, and obtain multiple product tags;
[0014] For each of the historical users, the static basic information and each of the product labels are constructed into a static sample; wherein the sample label of the static sample is determined based on the historical purchase information, and the sample label indicates that the historical user has purchased the product corresponding to the product label, or that the historical user has not purchased the product corresponding to the product label;
[0015] The first initial model is trained based on the multiple static samples to obtain a first recommendation model.
[0016] In one embodiment, the training process of the second recommendation model includes:
[0017] Obtain static basic information and historical behavior information of multiple historical users, and obtain multiple product tags;
[0018] For each of the historical users, the static basic information of the historical user and each of the product labels are constructed into a behavior sample; the sample label of the behavior sample is determined based on the behavior information of the historical user, and the sample label represents the positive or negative behavior of the historical user on the product corresponding to the product label;
[0019] The second initial model is trained based on the plurality of behavior samples to obtain a second recommendation model.
[0020] In one embodiment, the risk prediction model training process includes:
[0021] Obtain static basic information and historical risk information of multiple historical users, wherein the historical risk information includes at least one of historical claim settlement information, historical policy cancellation information, and historical disease information;
[0022] For each of the historical users, the static basic information and each item of the historical risk information are constructed into a risk sample, and the product label corresponding to the historical risk information is used as a sample label;
[0023] The third initial model is trained based on the multiple risk samples to obtain the risk prediction model.
[0024] In one embodiment, the screening of the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list includes:
[0025] For each of the first recommendation tags, based on the risk probability of each of the risk product tags in the risk list and the first recommendation probability of the first recommendation tag, determine the comprehensive recommendation probability of the first recommendation tag; wherein, the first recommendation list also includes the first recommendation probability of each of the first recommendation tags, and the first recommendation probability is the probability that the product corresponding to the first recommendation tag matches the target user; the risk list also includes the risk probability corresponding to each of the risk product tags, and the risk probability is the probability that the target user will make a claim and / or surrender the product corresponding to the risk product tag;
[0026] For each of the second recommended tags, based on the risk probability of each of the risk product tags in the risk list and the second recommendation probability of the second recommended tags, determine a comprehensive recommendation probability of the second recommended tags; wherein the second recommendation list also includes the second recommendation probability of each of the second recommended tags, and the second recommendation probability is the probability that the product corresponding to the second recommended tag matches the target user;
[0027] Screening each of the first recommended tags and each of the second recommended tags based on the comprehensive recommendation probability, and determining a tag whose comprehensive recommendation probability is greater than a target threshold as a target recommended tag;
[0028] The target recommendation tags are sorted according to the comprehensive recommendation probability to obtain a target recommendation list.
[0029] In one embodiment, recommending products to the target user based on the target recommendation list includes:
[0030] Adjusting the target recommendation list based on the recommendation rule to obtain the adjusted target recommendation list;
[0031] Based on the adjusted target recommendation list, a recommended product for the target user is determined.
[0032] In a second aspect, the present application further provides a product recommendation device, the device comprising a first recommendation module, a second recommendation module, a risk prediction module, a comprehensive screening module and a recommended product determination module, wherein:
[0033] A first recommendation module, configured to process static basic information of a target user using a first recommendation model to obtain a first recommendation list including a plurality of first recommendation tags;
[0034] A second recommendation module, configured to process the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0035] A risk prediction module, used to process the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label;
[0036] a comprehensive screening module, configured to screen the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag;
[0037] A recommended product determination module is used to recommend products to the target user based on the target recommendation list.
[0038] In a third aspect, the present application further provides a server, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0039] Using the first recommendation model to process the static basic information of the target user, obtaining a first recommendation list including a plurality of first recommendation tags;
[0040] Processing the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0041] Processing the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label;
[0042] Filtering the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag;
[0043] Recommend products to the target user based on the target recommendation list.
[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:
[0045] Using the first recommendation model to process the static basic information of the target user, obtaining a first recommendation list including a plurality of first recommendation tags;
[0046] Processing the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0047] Processing the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label;
[0048] Filtering the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag;
[0049] Recommend products to the target user based on the target recommendation list.
[0050] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:
[0051] Using the first recommendation model to process the static basic information of the target user, obtaining a first recommendation list including a plurality of first recommendation tags;
[0052] Processing the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0053] Processing the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label;
[0054] Filtering the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag;
[0055] Recommend products to the target user based on the target recommendation list.
[0056] The above-mentioned product recommendation method, device, server and medium obtain a first recommendation list for the target user through the static basic information of the target user and the first recommendation model, and obtain a second recommendation list for the target user through the dynamic behavior information of the target user and the second recommendation model, and also obtain a risk product list for the target user based on the risk information of the target user and the risk prediction model; the first recommendation list represents the target user's basic needs for products, which may be relatively broad and have a large range; the second recommendation list can more accurately represent the target user's current (recent) product needs; the risk list represents the probability that the target user will incur risks when purchasing each product; therefore, a comprehensive screening is performed in combination with the first recommendation list, the second recommendation list and the risk list, and then a target recommendation label that can represent the target user's needs and has a relatively low risk probability is obtained, thereby improving the accuracy of the recommended products. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0058] Figure 1 An application environment diagram of a product recommendation method in an embodiment;
[0059] Figure 2 A schematic diagram of a process of a product recommendation method in one embodiment;
[0060] Figure 3 A schematic diagram of a flow chart of a step of determining a target recommendation list in one embodiment;
[0061] Figure 4 A flowchart of a method for determining a recommended product in one embodiment;
[0062] Figure 5 Schematic diagram of the training process of the first recommendation model in one embodiment;
[0063] Figure 6 is a schematic diagram of a training process of a second recommendation model in one embodiment;
[0064] Figure 7 A schematic diagram of a training process of a risk prediction model in one embodiment;
[0065] Figure 8 is a schematic diagram of the structure of a target platform in an embodiment;
[0066] Fig. 9 is a schematic diagram of the structure of a data layer in one embodiment;
[0067] Fig.10 is a structural block diagram of a product recommendation device in one embodiment;
[0068] Fig.11 FIG. 4 is a diagram showing the internal structure of a server in one embodiment. DETAILED DESCRIPTION
[0069] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0070] The product method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, consumers can communicate with server 104 through the network through terminal 102, and server 104 can actively send recommended products to terminal 102 for display, or server 104 responds to the recommendation request sent by the consumer to server 104 through terminal 102, so that server 104 sends recommended products to terminal 102 for display. The data storage system can store data that server 104 needs to process. The data storage system can be integrated on server 104, or it can be placed on the cloud or other network servers. Among them, terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.
[0071] In an exemplary embodiment, Figure 2 As shown, a product recommendation method is provided, which is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 21 to 25, wherein:
[0072] Step 21: Use the first recommendation model to process the static basic information of the target user to obtain a first recommendation list including multiple first recommendation tags.
[0073] For the embodiments of the present application, the static basic information is the attribute information of the user, which may include elements such as the user's age, gender, place of residence, annual income, and education level. The static attribute information of the user may be obtained by the server and stored in the data storage system when the user registers on the target platform. The specific elements of the static attribute information are not specifically limited in the embodiments of the present application. In one possible manner, the static attribute information may also include static attribute information of the user's family members, and may also include elements such as the user's medical history.
[0074] The target user may be a user who actively sends a recommendation request to the server through a terminal. When the server receives the recommendation request, it determines the user as the target user based on the user's identity in the recommendation request. The target user may also be a user determined based on screening information, which may be input by the staff of the target platform. For example, the screening information may be male users under the age of 35. In this case, a male user aged 28 is the target user.
[0075] The first recommendation model is pre-trained and can make predictions based on static basic information of any user, thereby obtaining a first recommendation list for the user; wherein the first recommendation list includes multiple first recommendation tags and a first recommendation probability corresponding to each first recommendation tag.
[0076] Specifically, each product of the target platform is divided in advance to obtain a product list including multiple product tags. The product tag represents the attributes of the product, and one product corresponds to at least one product tag. The first recommended tag is a product tag, and the first recommendation probability is the probability that the product corresponding to the first recommended tag matches the target user.
[0077] Step 22: Use the second recommendation model to process the dynamic behavior information of the target user to obtain a second recommendation list including multiple second recommendation tags.
[0078] For the embodiment of the present application, the dynamic behavior information includes multiple historical dynamic behaviors of the user, and the dynamic behavior includes the operations and operation time generated by the user on the products in the target platform; wherein the operations include likes, favorites, forwarding, and browsing for more than a preset time. In one example, a dynamic behavior can be: Favorites-A product label-October 8, 2023 16:20. When the user uses the APP (Application), mini-program, official account and web interface associated with the target platform through the server, the dynamic behavior generated by the user will be acquired by the relevant behavior detection component and sent to the server, and stored in the data storage system.
[0079] Specifically, the second recommendation model may process all dynamic behavior information of the target user, dynamic behavior information within a specified period, or dynamic behavior information within a preset length from the current moment, which is not specifically limited in the embodiments of the present application.
[0080] The second recommendation model is pre-trained. The second recommendation model can predict the product tags that the user may be interested in based on the dynamic behavior information of any user to obtain a second recommendation list. The second recommendation list includes multiple second recommendation tags and also includes the second recommendation probability of each second recommendation tag; the second recommendation probability is the probability that the product corresponding to the second recommendation tag matches the target user. The second recommendation tag is a product tag. Since the second recommendation list is predicted based on the dynamic behavior information of the target user, the second recommendation list can more accurately represent the user's current product needs.
[0081] Step 23: Use the risk prediction model to process the risk information of the target user to obtain a risk list, which includes at least one risk product label.
[0082] For the embodiment of the present application, the risk prediction model is pre-trained and can predict the historical risk information of any user, so as to obtain a risk list for the user. Among them, the risk list includes at least one risk product label, and also includes the risk probability corresponding to each risk product label, wherein the risk probability represents the probability that the user may cancel the insurance and / or cancel the order when purchasing the product corresponding to the risk product label. The risk information includes at least one of the user's historical cancellation information, claim information and disease information; and each piece of information corresponds to a corresponding occurrence time. In the case where the target platform is an insurance business platform, since the relevant insurance products have periodic properties, associating the corresponding time with each piece of information / data can improve the accuracy of information processing and prediction.
[0083] Step 24: Screen the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, where the target recommendation list includes at least one target recommendation tag.
[0084] For the embodiment of the present application, the first recommendation list represents the basic product needs of the target user, which may be relatively broad and wide in scope; the second recommendation list can more accurately represent the current (recent) product needs of the target user; the risk list represents the probability that the target user will incur risks when purchasing each product; therefore, a comprehensive screening is performed in combination with the first recommendation list, the second recommendation list and the risk list, thereby obtaining a target recommendation label that can represent the needs of the target user and has a relatively low risk probability.
[0085] Step 25: Recommend products to target users based on the target recommendation list.
[0086] For the embodiments of the present application, the target recommendation tag can be a product tag of a specific product or a tag of a certain category of products; therefore, when the target recommendation tag is a specific product tag, the corresponding recommended product is determined directly based on the target recommendation tag; when the target recommendation tag is a tag of a certain category of products, based on the priority of each product corresponding to the tag of this category of products, the recommended product is determined from the products corresponding to the tag of this category of products; further, the recommended product for the target user is determined based on the target product tag, and then the recommended product is recommended to the target user.
[0087] In the above product recommendation method, a first recommendation list for the target user is obtained through the static basic information of the target user and the first recommendation model, and a second recommendation list for the target user is obtained through the dynamic behavior information of the target user and the second recommendation model, and a risk product list for the target user is obtained based on the risk information of the target user and the risk prediction model; the first recommendation list represents the basic product needs of the target user, which may be relatively broad and have a large range; the second recommendation list can more accurately represent the current (recent) product needs of the target user; the risk list represents the probability that the target user will incur risks when purchasing each product; therefore, a comprehensive screening is performed in combination with the first recommendation list, the second recommendation list and the risk list, and then a target recommendation label that can represent the needs of the target user and has a relatively low risk probability is obtained, thereby improving the accuracy of the recommended products.
[0088] In one embodiment, referring to Figure 3 , step 24 may specifically include steps 241 to 244, wherein:
[0089] Step 241: For each first recommended tag, determine the comprehensive recommendation probability of the first recommended tag based on the risk probability of each risk product tag in the risk list and the first recommendation probability of the first recommended tag.
[0090] Specifically, if the first recommendation tag and the risk tag correspond to the same product tag, the comprehensive recommendation probability of this product tag is P=P1(1-P2); where P1 is the first recommendation probability and P2 is the risk probability. If the product tag corresponding to the first recommendation tag does not have a corresponding risk tag in the risk list, the first recommendation probability P1 of the first recommendation tag is its comprehensive recommendation probability P.
[0091] In one example, the first recommendation list includes three first recommendation tags, A1, A2, and A3, and the first recommendation probabilities corresponding to these three first recommendation tags are 70%, 50%, and 30% respectively; there are two risk tags, A1 and A2, in the risk list, and the corresponding risk probabilities are 90% and 50% respectively. In the first recommendation list, the comprehensive recommendation probabilities corresponding to A1, A2, and A3 are 7%, 25%, and 30% respectively.
[0092] Step 242: For each second recommended tag, determine the comprehensive recommendation probability of the second recommended tag based on the risk probability of each risk product tag in the risk list and the second recommendation probability of the second recommended tag.
[0093] Specifically, if the second recommendation tag and the risk tag correspond to the same product tag, the comprehensive recommendation probability of this product tag is P=P3(1-P2); where P3 is the second recommendation probability and P2 is the risk probability. If the product tag corresponding to the second recommendation tag does not have a corresponding risk tag in the risk list, the first recommendation probability P3 of the second recommendation tag is its comprehensive recommendation probability P.
[0094] In one example, the first recommendation list includes three first recommendation tags, B1, B2, and B3, and the first recommendation probabilities corresponding to these three first recommendation tags are 40%, 60%, and 80% respectively; there are two risk tags, A1 and A2, in the risk list, and the corresponding risk probabilities are 90% and 50% respectively. In the first recommendation list, the comprehensive recommendation probabilities corresponding to B1, B2, and B3 are 4%, 30%, and 80% respectively.
[0095] It should be noted that step 241 and step 242 can be executed sequentially in the order of numbering, or can be executed simultaneously, which is not specifically limited in the embodiments of the present application.
[0096] Step 243: Screen each first recommended tag and each second recommended tag based on the comprehensive recommendation probability, and determine the tag with a comprehensive recommendation probability greater than a target threshold as a target recommended tag;
[0097] Step 244: Sort the target recommendation tags according to the comprehensive recommendation probability to obtain a target recommendation list.
[0098] Specifically, each first recommendation tag and each second recommendation tag are uniformly sorted in order of comprehensive recommendation probability from high to low; wherein, the target recommendation tag can be a tag with a comprehensive recommendation probability in the top N%, or it can be the top M tags, which is determined by the actual recommendation rules, and is not specifically limited in the embodiments of the present application. Then, each target recommendation tag is further sorted from high to low according to the comprehensive recommendation probability to obtain a target recommendation list, wherein the comprehensive recommendation probability is represented as the priority of the target user recommendation, and the higher the comprehensive recommendation probability, the higher the priority recommended to the target user.
[0099] Further, refer to Figure 4 The product recommendation method of the embodiment of the present application may also include a rule recommendation process, wherein, after step 25, it also includes steps 26 and 27, wherein:
[0100] Step 26: adjusting the target recommendation list based on the recommendation rule to obtain an adjusted target recommendation list;
[0101] Step 27: Based on the adjusted target recommendation list, determine the recommended products for the target users.
[0102] Specifically, the recommendation rules can be set by the user, specifically, by the staff of the target platform; the recommendation rules may include limiting the recommendation priority of each target recommendation tag in the target recommendation list, and may also include deleting each target recommendation tag in the target recommendation list, and adding additional product tags as target recommendation tags. Recommendation rules can be formulated based on the actual needs of the market and platform business. Specifically, after step 25, if the execution conditions of the rule recommendation are met, steps 26 and 27 of the rule recommendation can be executed to recommend the products corresponding to each target recommendation tag in the adjusted target recommendation list to the target user. When the execution conditions of the rule recommendation are not met, the products corresponding to each target recommendation tag in the target recommendation list are directly recommended to the target user. Whether the execution conditions of the rule recommendation are met can be set by the user of the target platform.
[0103] Furthermore, in the following content, the training processes of the first recommendation model, the second recommendation model and the risk prediction model are respectively explained.
[0104] In one embodiment, referring to Figure 5 The training process of the first recommendation model may include steps 011 to 013, wherein:
[0105] Step 011: Obtain static basic information and historical purchase information of multiple historical users, and obtain multiple product tags;
[0106] Step 012: For each historical user, the static basic information and each product label are constructed into a static sample; wherein the sample label of the static sample is determined based on the historical purchase information, and the sample label indicates that the historical user has purchased the product corresponding to the product label, or the historical user has not purchased the product corresponding to the product label;
[0107] 013. Train the first initial model based on multiple static samples to obtain a first recommendation model.
[0108] Specifically, the static information and purchase information of historical users are stored in the data storage system of the server and can be directly retrieved. The purchase information includes each product purchased by the historical user and the purchase time. The multiple product tags are product tags in the pre-divided product list.
[0109] The number of static samples that can be generated by each user's static information and purchase information is the same as the number of product labels in the product list. For a historical user, a set of training data is constructed with the user's static basic information and each product label. The training sample corresponding to each set of training data is whether the user has purchased the product corresponding to the product label; the sample label can be represented by 1 and 0, where 1 indicates that the user has purchased the product, and 0 indicates that the user has not purchased the product. Among them, the training sample with label 1 is a positive sample, and the training sample with label 0 is a negative sample.
[0110] In one example, the static basic information of a historical user is Q, and the purchase information indicates that the user has purchased three products, A1, B1, and C1; there are five product labels in the product list, namely A1, B1, C1, D1, and E1. Then the user can generate 5 static samples, namely: Q-A1-1, Q-B1-1, Q-C1-1, Q-D1-0, and Q-E1-0. Among them, the sample data of each static sample also includes the purchase time, which is not shown in this example.
[0111] Furthermore, the first initial model is trained with each constructed static sample, so as to obtain a first recommendation model after satisfying corresponding training conditions.
[0112] In one embodiment, referring to Figure 6 The training process of the second recommendation model may include steps 021 to 023, wherein:
[0113] Step 021: Obtain static basic information and historical behavior information of multiple historical users, and obtain multiple product tags;
[0114] Step 022: for each historical user, the static basic information of the historical user and each product label are constructed into a behavior sample; the sample label of the behavior sample is determined based on the behavior information of the historical user, and the sample label represents the positive or negative behavior of the historical user on the product corresponding to the product label;
[0115] Step 023: Train the second initial model based on multiple behavior samples to obtain a second recommendation model.
[0116] Specifically, the historical behavior information of multiple historical users is retrieved from the data storage system of the server, where the historical behavior information can be all dynamic behavior information of the target user, dynamic behavior information within a specified period, or dynamic behavior information within a preset distance from the current moment. The historical behavior includes the operations generated by the user on the products in the target platform and the operation time.
[0117] The number of dynamic samples that can be generated by each user's historical behavior information is the same as the number of historical behaviors of the user's historical behavior information. For a historical user, the user's static basic information and each historical row of the user are used as a set of sample data. The sample label of each set of sample data is related to the positivity of the operation corresponding to the historical behavior in the set of sample data. Among them, liking, collecting, browsing for more than a preset time, and forwarding a product are all positive operations, and the sample label corresponding to the positive operation is 1; deleting, clearing, blocking, etc. on a product are all negative operations, and the sample label of the negative operation is 0. Among them, the dynamic sample with a label of 0 is a negative sample, and the dynamic sample with a label of 1 is a positive sample.
[0118] In one example, the static basic information of a historical user is Q, and there are 5 historical behaviors in its historical behavior information. These 5 historical behaviors are respectively for 5 products (product labels are A1, B1, C1, D1 and E1). The operations in the 5 historical behaviors are like, favorite, clear, block and share. Based on this historical user, 5 dynamic samples will be constructed, namely Q-A1-1, Q-B1-1, Q-C1-0, Q-D1-0 and Q-E1-1. Among them, the sample data of each dynamic sample also includes the operation time, which is not shown in this example.
[0119] Furthermore, the second initial model is trained with each constructed dynamic sample, so as to obtain a second recommendation model after satisfying corresponding training conditions.
[0120] In one embodiment, referring to Figure 7 The training process of the risk prediction model may include steps 031 to 033, wherein:
[0121] Step 031: Obtain static basic information and historical risk information of multiple historical users, where the historical risk information includes at least one of historical claim settlement information, historical policy cancellation information, and historical disease information;
[0122] Step 032: For each historical user, construct a risk sample by combining the static basic information and each item of historical risk information, and use the product label corresponding to the historical risk information as the sample label;
[0123] Step 033: Train the third initial model based on multiple risk samples to obtain a risk prediction model.
[0124] Specifically, historical risk information of multiple historical users is retrieved from the data storage system of the server. The historical risk information may include at least one of historical claim information, historical policy cancellation information, and historical disease information, and also includes the generation time corresponding to each piece of information.
[0125] The number of risk samples that can be generated by each user's historical risk information is the same as the number of historical risk items of the user; each historical risk item includes the product of the cancellation / claim, the reason for the cancellation / claim and the generation time; historical risk items can also include disease types and possible claim products corresponding to the disease type.
[0126] For a historical user, the user's basic information and each product label are used as sample data of a set of risk samples. The sample label is whether the user has a historical risk item for the product corresponding to the product label. The sample label can be represented by 1 and 0, where 1 indicates that the user has a historical risk item for the product corresponding to the product label, and 0 indicates that the user does not have a historical risk item for the product corresponding to the product label.
[0127] In an example, the static basic information of a historical user is Q, and there are three historical risk items in his historical behavior information, with product labels A1, B1, C1, D1 and E1 respectively; the product labels targeted by the three historical risk items are B1, C1 and D1 respectively, then five risk samples will be constructed based on this historical user, namely Q-A1-0, Q-B1-1, Q-C1-1, Q-D1-1 and Q-E1-0.
[0128] Furthermore, the third initial model is trained with each constructed risk sample, so as to obtain a risk prediction model after satisfying corresponding training conditions.
[0129] In one embodiment, the structure of the target platform constructed based on the product recommendation method of the present application is as follows: Figure 8 As shown, it specifically includes user portrait middle platform, content service middle platform, personalized recommendation system, third-party platform and APP platform.
[0130] As a bridge connecting the data center and the front-end operation touchpoints, the personalized recommendation system mainly includes two functions: rule recommendation and model recommendation. One is rule recommendation, including localized delivery, provincial selection, and differentiated delivery strategies, which can realize real-time configuration for active or passive product recommendations to users. The other is intelligent recommendation, that is, using the first recommendation model, the second recommendation model, and the risk prediction model in the embodiment of this application to actively or passively recommend products to users.
[0131] For rule recommendation, the solution of the embodiment of the present application relies on the rich user and product label system of the user portrait platform and the content service platform, combined with the pre-set business logic rules, to recommend content to users by selecting specific user groups or setting personalized content push rules. Its advantages are intuitive and easy to understand, easy to implement and adjust, able to quickly respond to specific business or policy needs, and easy to formulate and optimize recommendation strategies.
[0132] For model recommendation, the solution of the embodiment of the present application uses a deep learning algorithm to automatically learn and extract recommendation rules from a large amount of source data such as user purchases and user behaviors. Thanks to the ability of deep learning to capture complex data patterns, model recommendation can achieve more personalized and accurate recommendations. In addition, as the system is iterated and upgraded, the solution of the embodiment of the present application will continue to add new data features to the model to further improve the recommendation effect of the model recommendation.
[0133] In actual applications, personalized recommendation systems can combine specific application scenarios, data scale, business needs, and requirements for recommendation effects and explainability to achieve customized strategy recommendations. By combining the two major recommendation capabilities of rule and model recommendation, more flexible, accurate, and personalized recommendation effects can be achieved.
[0134] The early construction process of the target platform can be divided into data mining and feature engineering, modeling and optimization and upgrading, multi-model fusion, improving recommendation stability and personalization, and front-end multi-touchpoint empowerment. The following content explains each process in more detail.
[0135] Data mining and feature engineering: In the data mining stage, the feature data layer is as follows: Fig. 9 As shown in the figure, the target mining platform deeply mines and integrates information such as user basic attribute data and user behavior data, extracts static features such as user basic information, user protection gaps, user family characteristics and product feature data from a wide range of data, as well as risk features such as user medical data, claims data, and surrender data, and integrates dynamic features that reflect user interests and preferences in App searches, clicks, browsing, and purchases as multi-dimensional inputs for the recommendation system, achieving more accurate insights into user preferences, needs, and potential intentions. At the same time, product platform data was introduced, and research was conducted in cooperation with the actuarial department to conduct a comprehensive and in-depth analysis of the characteristics of insurance products, and successfully build a scientific product classification system that can reflect the core characteristics and differences of various insurance products. With the help of this classification system, the recommendation system can more accurately recommend insurance products that meet the personalized demands of users.
[0136] During the feature engineering stage, the target platform completes feature cleaning by processing missing values, outliers, and duplicate records to ensure data integrity and consistency. It also uses statistical methods such as the Pearson correlation coefficient and the Spearman rank correlation coefficient, as well as model-based methods (such as random forest feature importance) to quantify the correlation between features and identify redundant features and key features. At the same time, it uses a variety of strategies such as filtering, wrapping, and embedding to select the most predictive feature combination, thereby improving the interpretability and generalization ability of the model and completing feature screening.
[0137] Modeling and optimization upgrade: In the modeling stage, relying on the user portraits extracted by the previous feature engineering, we integrate a variety of cutting-edge machine learning algorithms to establish a first product recommendation model based on static features, a second recommendation model based on dynamic behavior features, and a risk prediction model based on user risk features. We then adapt them to the business scenarios and construct a target model matrix.
[0138] In terms of model evaluation, a comprehensive and detailed evaluation system is adopted, covering multiple dimensions such as precision, recall, F1 score, AUC-ROC curve, etc., to comprehensively measure the effectiveness of the recommendation system. In addition, an online evaluation link is introduced to further verify the practicality and adaptability of the model through real-time user feedback and interactive data. Combined with experimental designs such as A / B testing and cross-validation, the scheme of the embodiment of the present application can objectively compare the performance changes of different models or parameters before and after adjustment, ensuring the accuracy and reliability of the evaluation results. Using a combination of online and offline evaluation methods, the scheme of the embodiment of the present application can more comprehensively understand the actual effect of the recommendation system and provide a solid foundation for subsequent iterative optimization.
[0139] Based on the evaluation results, the model can be regularly monitored for areas for improvement and the direction of improvement can be identified. The recommendation model can be continuously iterated and optimized by introducing new data sources, optimizing feature processing, adjusting model architecture and parameters, adopting more advanced algorithm technologies, and adjusting strategies based on business feedback. At the same time, through real-time monitoring systems and feedback mechanisms, the performance of the model in the online environment is closely tracked to ensure that the recommendation system can quickly adapt to market fluctuations and changes in user behavior, and maintain its timeliness and competitiveness.
[0140] Multi-model fusion to improve recommendation stability and personalization: The target platform integrates three models: the first recommendation model based on static features, the second recommendation model based on dynamic behavior features, and the risk prediction model based on user risk features, capturing the diversity and risks of user interests from multiple angles. Combined with a variety of re-ranking strategies, such as disease risk adjustment, surrender risk adjustment, and company business rhythm adjustment, the multi-dimensional model prediction results of the model ranking layer are further integrated. This multi-model integration method not only significantly improves the stability and accuracy of the recommendation system prediction, but also brings users a more personalized recommendation experience that fits their needs and risk tolerance.
[0141] By adopting an integrated learning strategy, we ensure the diversity and complementarity of the models, so that each model can give full play to its advantages. Through the designed integrated algorithm, we can effectively integrate the model prediction results to achieve information complementarity and performance improvement. This strategy enhances the comprehensiveness and flexibility of the target platform, so as to better adapt to complex user needs and market environment, and provide rich and accurate recommendation content.
[0142] In summary, the product recommendation method of the embodiment of the present application has at least the following beneficial effects:
[0143] 1. Enhanced personalized recommendations: By integrating the user static tag recommendation model and the user dynamic behavior recommendation model, the system can provide more accurate and personalized insurance product recommendations. This solves the problem that the existing system only relies on basic user information for recommendations, and improves the accuracy of recommendations and user satisfaction.
[0144] 2. Reduce the risk of the recommendation system: Introducing a risk model based on user medical, claims, and policy cancellation data, combined with a machine learning algorithm, can more accurately assess the risk characteristics of users. This overcomes the risk issues brought by traditional recommendation systems and reduces the overall risk of the recommendation system.
[0145] 3. Connecting data from multiple systems: Using intelligent data integration methods to ensure that data from different sources can be effectively integrated and utilized, providing a more comprehensive user portrait. This solves the problem of data silos, improves data utilization efficiency, and enables the recommendation system to understand user needs from a more comprehensive perspective.
[0146] 4. Real-time data analysis: With real-time data analysis capabilities, it can promptly reflect the latest trends of market changes and user needs, and quickly adjust recommendation strategies. This overcomes the shortcomings of existing systems that rely on historical data for analysis, resulting in delayed recommendation results, and improves the timeliness and accuracy of recommendations.
[0147] 5. Increased explainability and transparency: The system design combines rule recommendation and model recommendation, which not only provides highly abstract model recommendations, but also provides technical means for rule recommendation for business departments. This solves the problem that it is difficult for highly abstract model recommendations to provide clear reasons for recommendations to operators, increases the credibility of the system, and facilitates the review of regulatory agencies to ensure the compliance and transparency of the system.
[0148] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0149] Based on the same inventive concept, the embodiment of the present application also provides a product recommendation device for implementing the product recommendation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more product recommendation device embodiments provided below can refer to the limitations on the product recommendation method above, and will not be repeated here.
[0150] In an exemplary embodiment, Fig.10 As shown, a product recommendation device is provided, including a first recommendation module 1001, a second recommendation module 1002, a risk prediction module 1003, a comprehensive screening module 1004 and a recommended product determination module 1005, wherein:
[0151] The first recommendation module 1001 is used to process the static basic information of the target user using the first recommendation model to obtain a first recommendation list including a plurality of first recommendation tags;
[0152] The second recommendation module 1002 is used to process the dynamic behavior information of the target user using the second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags;
[0153] The risk prediction module 1003 is used to process the risk information of the target user using the risk prediction model to obtain a risk list, which includes at least one risk product label;
[0154] A comprehensive screening module 1004 is used to screen the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, where the target recommendation list includes at least one target recommendation tag;
[0155] The recommended product determination module 1005 is used to recommend products to the target user based on the target recommendation list.
[0156] In the above-mentioned product recommendation device, a first recommendation list for the target user is obtained through the static basic information of the target user and the first recommendation model, and a second recommendation list for the target user is obtained through the dynamic behavior information of the target user and the second recommendation model, and a risk product list for the target user is also obtained based on the risk information of the target user and the risk prediction model; the first recommendation list represents the basic needs of the target user for products, which may be relatively broad and have a large range; the second recommendation list can more accurately represent the current (recent) product needs of the target user; the risk list represents the probability that the target user will incur risks when purchasing each product; therefore, a comprehensive screening is performed in combination with the first recommendation list, the second recommendation list and the risk list, and then a target recommendation label that can represent the needs of the target user and has a relatively low risk probability is obtained, thereby improving the accuracy of the recommended products.
[0157] In one embodiment, the product recommendation device further includes a training module, wherein the training module is specifically used for:
[0158] Obtain static basic information and historical purchase information of multiple historical users, and obtain multiple product tags;
[0159] For each historical user, the static basic information and each product label are constructed into a static sample; wherein the sample label of the static sample is determined based on the historical purchase information, and the sample label represents that the historical user has purchased the product corresponding to the product label, or the historical user has not purchased the product corresponding to the product label;
[0160] The first initial model is trained based on multiple static samples to obtain a first recommendation model.
[0161] In one embodiment, the product recommendation device further includes a training module, wherein the training module is specifically used for:
[0162] Obtain static basic information and historical behavior information of multiple historical users, and obtain multiple product tags;
[0163] For each historical user, the static basic information of the historical user and each product label are constructed into a behavior sample; the sample label of the behavior sample is determined based on the behavior information of the historical user, and the sample label represents the positive or negative behavior of the historical user on the product corresponding to the product label;
[0164] The second initial model is trained based on multiple behavior samples to obtain a second recommendation model.
[0165] In one embodiment, the product recommendation device further includes a training module, wherein the training module is specifically used for:
[0166] Obtain static basic information and historical risk information of multiple historical users, where the historical risk information includes at least one of historical claim settlement information, historical policy cancellation information, and historical disease information;
[0167] For each historical user, the static basic information and each item of historical risk information are constructed into a risk sample, and the product label corresponding to the historical risk information is used as the sample label;
[0168] The third initial model is trained based on multiple risk samples to obtain a risk prediction model.
[0169] In one embodiment, the comprehensive screening module 1004 is specifically used to:
[0170] For each first recommendation tag, based on the risk probability of each risk product tag in the risk list and the first recommendation probability of the first recommendation tag, determine the comprehensive recommendation probability of the first recommendation tag; wherein the first recommendation list also includes the first recommendation probability of each first recommendation tag, and the first recommendation probability is the probability that the product corresponding to the first recommendation tag matches the target user; the risk list also includes the risk probability corresponding to each risk product tag, and the risk probability is the probability that the target user will make a claim and / or surrender the product corresponding to the risk product tag;
[0171] For each second recommendation tag, based on the risk probability of each risk product tag in the risk list and the second recommendation probability of the second recommendation tag, determine the comprehensive recommendation probability of the second recommendation tag; wherein the second recommendation list also includes the second recommendation probability of each second recommendation tag, and the second recommendation probability is the probability that the product corresponding to the second recommendation tag matches the target user;
[0172] Screening each first recommendation table tag and each second recommendation tag based on the comprehensive recommendation probability, and determining a tag whose comprehensive recommendation probability is greater than a target threshold as a target recommendation tag;
[0173] Sort the target recommendation tags according to the comprehensive recommendation probability to obtain a target recommendation list.
[0174] In one embodiment, the recommended product determination module 1005 is specifically configured to:
[0175] Adjusting the target recommendation list based on the recommendation rule to obtain an adjusted target recommendation list;
[0176] Based on the adjusted target recommendation list, determine the recommended products for the target users.
[0177] Each module in the above-mentioned product recommendation device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of the processor in the server in the form of hardware, or can be stored in the memory in the server in the form of software, so that the processor can call and execute the corresponding operations of each module.
[0178] In an exemplary embodiment, a server is provided. The server may be a terminal, and its internal structure diagram may be as follows: Fig.11As shown. The server includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the server is used to provide computing and control capabilities. The memory of the server includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the server is used to exchange information between the processor and an external device. The communication interface of the server is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a product recommendation method is implemented. The display unit of the server is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the server can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the server shell, or an external keyboard, touchpad or mouse.
[0179] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the server to which the solution of the present application is applied. The specific server may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
[0180] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, any step in the above-mentioned product recommendation method embodiment is implemented.
[0181] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements any step in the above-mentioned product recommendation method embodiment. 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 this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant regulations.
[0182] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0183] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0184] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A product recommendation method, characterized in that: The method comprises: Using the first recommendation model to process the static basic information of the target user, obtaining a first recommendation list including a plurality of first recommendation tags; Processing the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags; Processing the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label; Filtering the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag; Recommend products to the target user based on the target recommendation list.
2. The method according to claim 1, characterized in that The training process of the first recommendation model includes: Obtain static basic information and historical purchase information of multiple historical users, and obtain multiple product tags; For each of the historical users, the static basic information and each of the product labels are constructed into a static sample; wherein the sample label of the static sample is determined based on the historical purchase information, and the sample label indicates that the historical user has purchased the product corresponding to the product label, or that the historical user has not purchased the product corresponding to the product label; The first initial model is trained based on the multiple static samples to obtain a first recommendation model.
3. The method according to claim 1, characterized in that The training process of the second recommendation model includes: Obtain static basic information and historical behavior information of multiple historical users, and obtain multiple product tags; For each of the historical users, the static basic information of the historical user and each of the product labels are constructed into a behavior sample; the sample label of the behavior sample is determined based on the behavior information of the historical user, and the sample label represents the positive or negative behavior of the historical user on the product corresponding to the product label; The second initial model is trained based on the plurality of behavior samples to obtain a second recommendation model.
4. The method according to claim 1, characterized in that: The training process of the risk prediction model includes: Acquire static basic information and historical risk information of multiple historical users, wherein the historical risk information includes at least one of historical claim settlement information, historical policy cancellation information, and historical disease information; For each of the historical users, the static basic information and each item of the historical risk information are constructed into a risk sample, and the product label corresponding to the historical risk information is used as a sample label; The third initial model is trained based on the multiple risk samples to obtain the risk prediction model.
5. The method according to any one of claims 1 to 4, characterized in that The screening of the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list includes: For each of the first recommendation tags, based on the risk probability of each of the risk product tags in the risk list and the first recommendation probability of the first recommendation tag, determine the comprehensive recommendation probability of the first recommendation tag; wherein, the first recommendation list also includes the first recommendation probability of each of the first recommendation tags, and the first recommendation probability is the probability that the product corresponding to the first recommendation tag matches the target user; the risk list also includes the risk probability corresponding to each of the risk product tags, and the risk probability is the probability that the target user will make a claim and / or surrender the product corresponding to the risk product tag; For each of the second recommended tags, based on the risk probability of each of the risk product tags in the risk list and the second recommendation probability of the second recommended tags, determine a comprehensive recommendation probability of the second recommended tags; wherein the second recommendation list also includes the second recommendation probability of each of the second recommended tags, and the second recommendation probability is the probability that the product corresponding to the second recommended tag matches the target user; Screening each of the first recommended tags and each of the second recommended tags based on the comprehensive recommendation probability, and determining a tag whose comprehensive recommendation probability is greater than a target threshold as a target recommended tag; The target recommendation tags are sorted according to the comprehensive recommendation probability to obtain a target recommendation list.
6. The method according to claim 5, characterized in that The recommending products to the target user based on the target recommendation list includes: Adjusting the target recommendation list based on the recommendation rule to obtain the adjusted target recommendation list; Based on the adjusted target recommendation list, a recommended product for the target user is determined.
7. A product recommendation device, characterized in that: The device includes a first recommendation module, a second recommendation module, a risk prediction module, a comprehensive screening module and a recommended product determination module, wherein: A first recommendation module, configured to process static basic information of a target user using a first recommendation model to obtain a first recommendation list including a plurality of first recommendation tags; A second recommendation module, configured to process the dynamic behavior information of the target user using a second recommendation model to obtain a second recommendation list including a plurality of second recommendation tags; A risk prediction module, used to process the risk information of the target user using a risk prediction model to obtain a risk list, wherein the risk list includes at least one risk product label; a comprehensive screening module, configured to screen the first recommendation list and the second recommendation list based on the risk list to obtain a target recommendation list, wherein the target recommendation list includes at least one target recommendation tag; A recommended product determination module is used to recommend products to the target user based on the target recommendation list.
8. A server, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.