Business Object Recommendation Method, Device, Computer Equipment and Storage Medium

Abstract

The present application relates to a business object recommendation method, apparatus, computer device, and storage medium. The method includes: in response to business recommendation requirement information of a resource interaction platform, obtaining user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information; respectively performing feature extraction on the user account information, business information, business interaction information, and business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; inputting the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; and performing business object recommendation for the user account based on the at least one business recommendation information corresponding to the business recommendation requirement information. By using this method, the accuracy and efficiency of business object recommendation for the user account can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and particularly to a method, apparatus, computer device, storage medium, and computer program product for recommending business objects. Background Art

[0002] With the development of computer technology, business recommendation technology has emerged. This technology has the characteristics that data stores information, algorithms provide logic, and the architecture liberates hands. Among them, data stores information, including the attributes of users and content, and the behavior preferences of users; algorithms provide logic. As data accumulates continuously, a huge amount of information is stored, and a complex information processing logic is required to return recommended content or services based on the logic.

[0003] In traditional technologies, there are two ways for financial institutions to recommend services to customers: manual recommendation and traditional collaborative filtering recommendation algorithms. The method of manual recommendation requires a large amount of time for people to understand the basic situation and preferences of users. At the same time, it ignores the fact that with the change of the market environment, the needs and preference behaviors of users for financial services will also be affected, and it cannot recommend services in a timely and accurate manner. Using traditional methods for service recommendation results in a low accuracy rate of the recommended services. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium, and computer program product for recommending business objects that can improve the accuracy rate of recommended services.

[0005] In a first aspect, the present application provides a method for recommending business objects. The method includes: in response to the business recommendation requirement information of a resource interaction platform, obtaining the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information; respectively performing feature extraction on the user account information, the business information, the business interaction information, and the business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; inputting the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through sample data corresponding to at least one piece of the business recommendation requirement information; based on at least one business recommendation information corresponding to the business recommendation requirement information, performing business object recommendation for the user account.

[0006] In a second aspect, the present application also provides a business object recommendation device. The device includes: an information acquisition module, configured to acquire user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information in response to the business recommendation requirement information of the resource interaction platform; a feature extraction module, configured to perform feature extraction on the user account information, the business information, the business interaction information, and the business environment information respectively, to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; an information processing module, configured to input the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model, to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained by sample data corresponding to at least one business recommendation requirement information; a business recommendation module, configured to perform business object recommendation for the user account based on at least one business recommendation information corresponding to the business recommendation requirement information.

[0007] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: acquiring user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information in response to the business recommendation requirement information of the resource interaction platform; performing feature extraction on the user account information, the business information, the business interaction information, and the business environment information respectively, to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; inputting the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model, to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained by sample data corresponding to at least one business recommendation requirement information; performing business object recommendation for the user account based on at least one business recommendation information corresponding to the business recommendation requirement information.

[0008] Fourth aspect, the present application also provides a computer-readable storage medium. On the computer-readable storage medium, there is a computer program stored, and when the computer program is executed by a processor, the following steps are implemented: in response to the business recommendation requirement information of the resource interaction platform, obtain the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information; respectively perform feature extraction on the user account information, the business information, the business interaction information, and the business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; input the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through sample data corresponding to at least one of the business recommendation requirement information; based on the at least one business recommendation information corresponding to the business recommendation requirement information, perform business object recommendation for the user account.

[0009] Fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented: in response to the business recommendation requirement information of the resource interaction platform, obtain the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information; respectively perform feature extraction on the user account information, the business information, the business interaction information, and the business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; input the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through sample data corresponding to at least one of the business recommendation requirement information; based on the at least one business recommendation information corresponding to the business recommendation requirement information, perform business object recommendation for the user account.

[0010] The above-mentioned method, device, computer equipment, storage medium and computer program product for business object recommendation obtain user account information, business information, business interaction information and business environment information corresponding to the business recommendation requirement information in response to the business recommendation requirement information of the resource interaction platform; respectively extract features from the user account information, business information, business interaction information and business environment information to obtain a user feature vector, a business feature vector and a business environment feature vector corresponding to the business recommendation requirement; input the user feature vector, the business feature vector and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through sample data corresponding to at least one business recommendation requirement information; based on at least one business recommendation information corresponding to the business recommendation requirement information, business object recommendation is performed on the user account.

[0011] By perceiving the business environment information in real time, the user account information, business information, business interaction information and business environment information are input into a business recommendation model pre-constructed to conform to the current business environment; through the business recommendation model, the recommendation probability of the user for each business is calculated, and in the face of a changing business environment scenario, accurate recommendation of business objects can be made according to the current business environment and the historical interaction situation between the user account and the business; at the same time, for the cold start problem of user accounts that have not used the corresponding business, more reasonable business object recommendations can also be provided for the user account through the user account, business situation and current business environment information. It can effectively improve the accuracy and efficiency of business object recommendation for user accounts. Brief Description of the Drawings

[0012] Figure 1 It is an application environment diagram of a method for business object recommendation in an embodiment;

[0013] Figure 2 It is a flowchart of a method for business object recommendation in an embodiment;

[0014] Figure 3 It is a flowchart of a method for obtaining business recommendation information in an embodiment;

[0015] Figure 4 It is a flowchart of a method for obtaining the first sub-business recommendation information in an embodiment;

[0016] Figure 5 It is a flowchart of a method for obtaining the second sub-business recommendation information in an embodiment;

[0017] Figure 6Schematic diagram of the process for obtaining the second sub-service recommendation information in another embodiment;

[0018] Figure 7 Schematic diagram of the process for obtaining the service recommendation information in another embodiment;

[0019] Figure 8 Schematic diagram of the process for obtaining the feature vector in one embodiment;

[0020] Figure 9 Schematic diagram of the structure of the pre-trained service recommendation model in one embodiment;

[0021] Figure 10 Schematic diagram of the implementation logic of a service object recommendation method in one embodiment;

[0022] Figure 11 Block diagram of the structure of a service object recommendation device in one embodiment;

[0023] Figure 12 Internal structure diagram of a computer device in one embodiment. Detailed implementation manners

[0024] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to 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.

[0025] A service object recommendation method provided by an embodiment of the present application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. In response to the business recommendation requirement information of the resource interaction platform, the server 104 obtains the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information from the terminal 102; respectively extracts features from the user account information, business information, business interaction information, and business environment information to obtain the user feature vector, business feature vector, and business environment feature vector corresponding to the business recommendation requirement; inputs the user feature vector, business feature vector, and business environment feature vector into the pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained by the sample data corresponding to at least one business recommendation requirement information; based on at least one business recommendation information corresponding to the business recommendation requirement information, business object recommendations are made for the user account. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0026] In one embodiment, as Figure 2 shown, a business object recommendation method is provided. Taking the server in Figure 1 as an example, the method includes the following steps:

[0027] Step 202, in response to the business recommendation requirement information of the resource interaction platform, obtain the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information.

[0028] Among them, the resource interaction platform can be a platform required for users to handle business using user accounts and cause resource scheduling corresponding to the business. For example: banks, exchanges, fund companies, etc.

[0029] Among them, the business recommendation requirement information can be the requirement corresponding to recommending a business that meets the actual situation to the user account from the resource interaction platform according to the actual information of the user's user account. For example: Account A has a large amount of funds and often buys stocks from the exchange. Then, according to the actual information of Account A, the requirement of recommending suitable stocks to Account A.

[0030] Among them, the user account information can be the basic information, inherent information, etc. carried by the user corresponding to the business recommendation.

[0031] Among them, the business information can be the basic information, inherent information, etc. carried by each business in the resource interaction platform.

[0032] Among them, the business interaction information can be the information generated by which businesses the user account has interacted with.

[0033] Among them, the business environment information can be the historical interaction data of each business in the resource interaction platform in the resource interaction platform, and the market environment data corresponding to each business. Among them, the market environment data can be the data reflecting the current social situation.

[0034] Specifically, in response to the business recommendation requirement information of the resource interaction platform, the server obtains the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information from the terminal according to the instruction of the server response terminal, and stores the obtained user account information, business information, business interaction information, and business environment information in the storage unit. For example, the user account information can be age, occupation, marital status, education level, whether there is a default record, average account balance, whether there is a loan, etc.; the business information can be business type, business risk, business income, business term, business security, etc.; the business interaction information can be which businesses the user account has historically purchased or handled; the business environment information can be the region of the user account, unemployment rate, consumption index, bank interest rate, etc. When the server needs to process any data record in the user account information, business information, business interaction information, and business environment information, it is retrieved from the storage unit to the volatile storage resource for the central processing unit to calculate. Among them, any data record can be a single data input to the central processing unit, or multiple data can be input to the central processing unit at the same time.

[0035] For example, the server 104 responds to the instruction of the terminal 102, obtains at least two historical account attribute information from the terminal 102, and stores them in the storage unit in the server 104. Among them, there are 10 data records corresponding to the inherent information obtained by the server 104, and multiple data can be input at the same time.

[0036] Step 204: Respectively extract features from the user account information, business information, business interaction information, and business environment information to obtain the user feature vector, business feature vector, and business environment feature vector corresponding to the business recommendation requirement.

[0037] Among them, the user feature vector can be a vector generated after extracting features related to the user account from the text information in the user account information, business information, business interaction information, and business environment information.

[0038] Among them, the business feature vector can be a vector generated after extracting features related to each business from the text information in the user account information, business information, business interaction information, and business environment information.

[0039] Among them, the business environment feature vector can be a multi-dimensional feature vector formed by extracting features from the non-text information in the user account information, business information, business interaction information, and business environment information.

[0040] Specifically, for the text information in the user account information, business information, business interaction information, and business environment information, construct an information dictionary mapping table corresponding to the business recommendation requirement information, and map it to a numerical value using the information dictionary mapping table. Among them, the data for constructing the information dictionary mapping table is the text information of the user account information, business information, and business interaction information, such as occupation, marital status, whether there is a default record, region, etc. Based on the information dictionary mapping table, then use One-hot encoding to extract features from each information and the mapping relationship between each information in the user account information, business information, business interaction information, and business environment information in the information dictionary mapping table to obtain the user feature vector and the business feature vector. Among them, the information dictionary mapping table is shown in Table 1.

[0041] Table 1: Information Dictionary Mapping Table

[0042]

[0043]

[0044] For the non-text information in the user account information, business information, business interaction information, and business environment information, after feature extraction, represent it in the form of constructing a multi-dimensional feature vector to obtain the business environment feature vector. Among them, the basic information processing of the business environment feature vector is shown in Table 2.

[0045] Table 2: Basic Information Processing of Business Environment Feature Vector

[0046] ID Occupation Marital Status Default Situation ... Unemployment Rate Consumer Price Index Interest Rate 1 Worker Unmarried No Default ... 0.5 100.9 0.4 1 0010 00 0 ... 0.5 100.9 0.4

[0047] Step 206, input the user feature vector, business feature vector, and business environment feature vector into the pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information.

[0048] Among them, the business recommendation model can be a market environment perception recommendation model. The market environment perception recommendation model is mainly composed of a Generalized Matrix Factorization (GMF) model and an Environment Joint Multilayer Perceptron (EMP) model.

[0049] Among them, the business recommendation information can be the probability information obtained by calculating the input quantity through the business recommendation model. The business recommendation information can represent the matching degree between the business and the user account.

[0050] Specifically, the user feature vector, the business feature vector, and the business environment feature vector are input into the pre-trained business recommendation model. The business recommendation model is mainly composed of a Generalized Matrix Factorization (GMF) model and an Environment Joint Multilayer Perceptron (EMP) model. The Generalized Matrix Factorization model is introduced to process the interaction information between the user account and the business. At the same time, the Environment Joint Multilayer Perceptron is introduced to fuse the market environment information and the information features of the user and the product. Finally, by scoring the results of the Generalized Matrix Factorization model and the Environment Joint Multilayer Perceptron model, the best result is obtained, that is, at least one business recommendation information corresponding to the business recommendation demand information. The structural schematic diagram of the pre-trained business recommendation model is as Figure 9 shown.

[0051] Regarding the calculation process of the Generalized Matrix Factorization model, the non-linearity of the activation function in the Generalized Matrix Factorization model is used to express the interaction relationship between the user account and the business. The formula is as follows:

[0052] φ GMF = out (h T (p u * i ))

[0053] Among them, φ GMF represents the prediction result of the Generalized Matrix Factorization model, p u and q i represent the user feature vector and the business feature vector respectively, and a out and h T represent the Sigmoid activation function of the output layer and the connection weight respectively.

[0054] Regarding the calculation process of the environment - joint multi - layer perceptron model, the environment - joint multi - layer perceptron model concatenates the business environment feature vector, user feature vector, and business feature vector, and then inputs them into the multi - layer perceptron. Through multiple - layer neural networks, it mines more complex and effective feature combinations, thereby integrating business environment information into the model. At the same time, it can capture the relationships among the business environment, user accounts, and businesses. The formula is as follows:

[0055]

[0056] Among them, φ EMP represents the prediction result of the environment - joint multi - layer perceptron model, c m represents the business environment feature vector, p u represents the user feature vector, q i represents the business feature vector, w i and b i and f represent the connection weights, biases, and activation functions of the model respectively.

[0057] Regarding the generalized matrix factorization model and the environment - joint multi - layer perceptron model in the business recommendation model, they are both trained through sample data corresponding to at least one business recommendation requirement information.

[0058] For the generalized matrix factorization model, when the business environment is relatively stable or there is a large amount of interaction information between the user account and the business, the model performs well; for the environment - joint multi - layer perceptron model, when the business environment changes greatly or there is less interaction information between the user account and the business, the model performs well. Therefore, by scoring the results of the generalized matrix factorization model and the environment - joint multi - layer perceptron model, a better and more comprehensive result can be obtained. The formula is as follows:

[0059]

[0060] Among them, φ GMF and φ EMP represent the output results of the generalized matrix factorization model and the environment - joint multi - layer perceptron model respectively. Their influence on the final result is adjusted through the connection weight h. σ represents the Sigmoid activation function. Through the processing of the activation function, the final result of the multi - model prediction is output between 0 and 1. The closer it is to 1, the closer the business result predicted by the model is to the true needs of the user account.

[0061] Step 208: Based on at least one business recommendation information corresponding to the business recommendation requirement information, recommend business objects to the user account.

[0062] Specifically, the business recommendation information is sorted in ascending order, and the business with business recommendation information greater than a preset threshold is selected as the business object for recommendation to the user account; wherein, the business object refers to the business that meets the attributes of the user account. A schematic diagram of the implementation logic of a business object recommendation method is as Figure 10 shown.

[0063] In the above-mentioned business object recommendation method, by responding to the business recommendation requirement information of the resource interaction platform, the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information are obtained; feature extraction is respectively performed on the user account information, business information, business interaction information, and business environment information to obtain the user feature vector, business feature vector, and business environment feature vector corresponding to the business recommendation requirement; the user feature vector, business feature vector, and business environment feature vector are input into the pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through the sample data corresponding to at least one business recommendation requirement information; based on at least one business recommendation information corresponding to the business recommendation requirement information, business object recommendation is performed on the user account.

[0064] By performing real-time perception on the business environment information, the user account information, business information, business interaction information, and business environment information are input into the business recommendation model pre-constructed to conform to the current business environment; the recommendation probability of the user for each business is calculated through the business recommendation model, and in the face of a changing business environment scenario, accurate recommendation of business objects can be performed according to the current business environment and the historical interaction situation between the user account and the business; at the same time, for the cold start problem of user accounts that have not used the corresponding business, more reasonable business object recommendations can also be provided to the user account through the user account, business situation, and current business environment information. It can effectively improve the accuracy and efficiency of business object recommendation for user accounts.

[0065] In one embodiment, as Figure 3 shown, inputting the user feature vector, business feature vector, and business environment feature vector into the pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information includes:

[0066] Step 302, inputting the user feature vector and the business feature vector into the matrix factorization model to obtain at least one first sub-business recommendation information corresponding to the business recommendation requirement information.

[0067] Among them, the matrix factorization model can be a Generalized Matrix Factorization (GMF).

[0068] Among them, the first sub-service recommendation information can be the probability obtained after calculation using the generalized matrix factorization model.

[0069] Specifically, input the user feature vector and the service feature vector into the generalized matrix factorization model. The non-linearity of the activation function in the generalized matrix factorization model can be used to express the interaction relationship between the user account and the service. The formula is as follows:

[0070] φ GMF = out (h T (p u * i ))

[0071] Among them, φ GMF represents the prediction result of the generalized matrix factorization model, p u and q i represent the user feature vector and the service feature vector respectively, a out and h T represent the Sigmoid activation function of the output layer and the connection weight respectively. Based on the calculation of the generalized matrix factorization model, at least one first sub-service recommendation information corresponding to the service recommendation requirement information is obtained.

[0072] Step 304, input the user feature vector, the service feature vector, and the service environment feature vector into the environment perception model to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

[0073] Among them, the environment perception model can be an Environment Joint Multilayer Perceptron (EMP).

[0074] Among them, the second sub-service recommendation information can be the probability obtained after calculation using the environment joint multilayer perceptron model.

[0075] Specifically, input the user feature vector, the service feature vector, and the service environment feature vector into the environment joint multilayer perceptron model. The environment joint multilayer perceptron model splices the service environment feature vector, the user feature vector, and the service feature vector, and then inputs them into the multilayer perceptron. Through multiple-layer neural networks, more complex and effective feature combinations are mined, so as to integrate the service environment information into the model. At the same time, the relationship between the service environment, the user account, and the service can be captured. The formula is as follows:

[0076]

[0077] Among them, φ EMP represents the prediction result of the environment joint multi-layer perceptron model, c m represents the business environment feature vector, p u represents the user feature vector, q i represents the business feature vector, w i , b i and f respectively represent the connection weight, bias and activation function of the model. Based on the calculation of the environment joint multi-layer perceptron model, at least one second sub-business recommendation information corresponding to the business recommendation requirement information is obtained.

[0078] Step 306, adjust each first sub-business recommendation information and each second sub-business recommendation information to obtain at least one business recommendation information corresponding to the business recommendation requirement information.

[0079] Specifically, when adjusting each first sub-business recommendation information and each second sub-business recommendation information, for the generalized matrix factorization model, the model performs well when the business environment is relatively stable or there is a large amount of interaction information between the user account and the business; while for the environment joint multi-layer perceptron model, the model performs well when the business environment changes greatly or there is less interaction information between the user account and the business. Therefore, scoring the results of the generalized matrix factorization model and the environment joint multi-layer perceptron model can obtain better and more comprehensive results. The formula is as follows:

[0080]

[0081] Among them, φ GMF , φ EMP respectively represent the output results of the generalized matrix factorization model and the environment joint multi-layer perceptron model, and their influence on the final result is adjusted by the connection weight h. σ represents the Sigmoid activation function. Through the processing of the activation function, the final result of the multi-model prediction is output between 0 and 1. The closer to 1, the closer the business result predicted by the model is to the true need tendency of the user account, that is, at least one business recommendation information corresponding to the business recommendation requirement information is obtained.

[0082] In this embodiment, by inputting the user feature vector, business feature vector and business environment feature vector into the matrix factorization model and environment perception model in the pre-trained business recommendation model respectively, the weight can be adjusted according to the stability of the business environment or the quantity of interaction information between the user account and the business, so as to improve the accuracy of the business object recommendation for the user account.

[0083] In one embodiment, such as Figure 4As shown, input the user feature vector and the service feature vector into the matrix factorization model to obtain at least one first sub-service recommendation information corresponding to the service recommendation requirement information, including:

[0084] Step 402: Multiply the user feature vector by the service feature vector to obtain the result of the dot product of the feature vectors.

[0085] Among them, the result of the dot product of the feature vectors can be the calculation result after multiplying the user feature vector by the service feature vector.

[0086] Specifically, multiply the user feature vector by the user feature vector according to the rules of vector dot product to obtain the result of the dot product of the feature vectors after dot product calculation. The calculation formula is as follows:

[0087] A = p u * i

[0088] Among them, p u and q i respectively represent the user feature vector and the service feature vector.

[0089] Step 404: Perform a first connection weight adjustment on the result of the dot product of the feature vectors to obtain the result of the first connection weight adjustment.

[0090] Among them, the result of the first connection weight adjustment can be the result obtained by adjusting the result of the dot product of the feature vectors using the connection weight; among them, the first connection weight adjustment can be one of the connection weights in the generalized matrix factorization model.

[0091] Specifically, use the first connection weight to act on the result of the dot product of the feature vectors and adjust the weight of the result of the dot product of the feature vectors to obtain the result of the first connection weight adjustment. The calculation formula is as follows:

[0092] B = h T (A) = h T (p u * i )

[0093] Among them, p u and q i respectively represent the user feature vector and the service feature vector, and h T respectively represent the connection weights of the output layer.

[0094] Step 406: Obtain at least one first sub-service recommendation information according to the result of the first connection weight adjustment and the first mapping relationship.

[0095] Among them, the first mapping relationship can be one of the Sigmoid activation functions of the output layer in the generalized matrix factorization model.

[0096] Specifically, by using the Sigmoid activation function of one of the output layers in the generalized matrix factorization model to calculate the first connection weight adjustment result, at least one first sub-service recommendation information can be obtained. The calculation formula is as follows:

[0097] φ GMF = out (h T (p u * i ))

[0098] Among them, φ GMF represents the prediction result of the generalized matrix factorization model, p u and q i represent the user feature vector and the service feature vector respectively, a out and h T represent the Sigmoid activation function of the output layer and the connection weight respectively. Based on the calculation of the generalized matrix factorization model, at least one first sub-service recommendation information corresponding to the service recommendation requirement information is obtained.

[0099] In this embodiment, by using the matrix factorization model to process the dot product result of the user feature vector and the service feature vector, the stronger learning and representation ability of the matrix factorization model can be utilized to make the calculated service recommendation probability of a part of the service recommendation requirement information more accurate.

[0100] In one embodiment, as Figure 5 shown, input the user feature vector, the service feature vector, and the service environment feature vector into the environment perception model to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information, including:

[0101] Step 502, splice the user feature vector, the service feature vector, and the service environment feature vector to obtain a feature vector splicing result.

[0102] Among them, the feature vector splicing result can be a numerical value or a vector obtained by splicing the user feature vector, the service feature vector, and the service environment feature vector.

[0103] Specifically, splice the user feature vector, the service feature vector, and the service environment feature vector to obtain a feature vector splicing result with the service environment feature vector in the first row, the user feature vector in the second row, and the service feature vector in the third row.

[0104] Step 504, input the feature vector splicing result into the recommendation information output layer to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

[0105] Among them, the recommended information output layer can be the functional layer of the environment - joint multi - layer perceptron model for calculating probabilities, and it can perform processing on connection weights, biases, and activation functions multiple times according to actual requirements.

[0106] Specifically, the concatenated result of the feature vectors is input into the recommended information output layer, that is, into the multi - layer perceptron. Through multiple - layer neural networks, more complex and effective feature combinations are mined, so that business environment information can be incorporated into the model. At the same time, the relationship between the business environment, user accounts, and the business can be captured. The formula is as follows:

[0107]

[0108] Among them, φ EMP represents the prediction result of the environment - joint multi - layer perceptron model, c m represents the business environment feature vector, p u represents the user feature vector, q i represents the business feature vector, w i , b i and f respectively represent the connection weight, bias, and activation function of the model. Based on the calculation of the environment - joint multi - layer perceptron model, at least one second sub - business recommended information corresponding to the business recommendation requirement information is obtained.

[0109] In this embodiment, by using the environment - perception model to calculate the concatenated result of the user feature vector, business feature vector, and business environment feature vector, and using the recommended information output layer of the environment - perception model to perform multiple probability calculation iterations on the concatenated result of the feature vectors, the relationship between the business environment, user accounts, and the business can be captured, the performance of the model can be improved, and the accuracy of the model can be increased.

[0110] In one embodiment, as Figure 6 shown, inputting the concatenated result of the feature vectors into the recommended information output layer to obtain at least one second sub - business recommended information corresponding to the business recommendation requirement information includes:

[0111] Step 602, perform a second connection weight adjustment on the concatenated result of the feature vectors to obtain a second connection weight adjustment result; adjust the second connection weight adjustment result to obtain a first adjustment result; according to the first adjustment result and the second mapping relationship, obtain at least one first intermediate business recommended information.

[0112] Among them, the second connection weight adjustment result can be the result obtained by adjusting the concatenated result of the feature vectors using the connection weight; among them, the second connection weight adjustment can be the first connection weight in the environment - joint multi - layer perceptron model.

[0113] Among them, the first adjustment result can be the result obtained by applying the first bias of the environment joint multi-layer perceptron model to the second connection weight adjustment result.

[0114] Among them, the second mapping relationship can be the Sigmoid activation function of the first output layer in the environment joint multi-layer perceptron model.

[0115] Among them, the first intermediate service recommendation information can be the result obtained by processing the first adjustment result through the first activation function of the environment joint multi-layer perceptron model.

[0116] Specifically, the connection weight of the first connection weight in the environment joint multi-layer perceptron model is used to adjust the weight of the feature vector splicing result to obtain the second connection weight adjustment result; then, the first bias of the environment joint multi-layer perceptron model is used to adjust the second connection weight adjustment result to obtain the first adjustment result; finally, the Sigmoid activation function of the first output layer in the environment joint multi-layer perceptron model is used to process the first adjustment result to obtain at least one first intermediate service recommendation information.

[0117] Step 604: Adjust the third connection weight for each first intermediate service recommendation information to obtain the third connection weight adjustment result; adjust the third connection weight adjustment result to obtain the second adjustment result; according to the second adjustment result and the third mapping relationship, obtain at least one second intermediate service recommendation information.

[0118] Among them, the third connection weight adjustment result can be the result obtained by adjusting each first intermediate service recommendation information using the connection weight; among them, the third connection weight adjustment can be the second connection weight in the environment joint multi-layer perceptron model.

[0119] Among them, the second adjustment result can be the result obtained by applying the second bias of the environment joint multi-layer perceptron model to the third connection weight adjustment result.

[0120] Among them, the third mapping relationship can be the Sigmoid activation function of the second output layer in the environment joint multi-layer perceptron model.

[0121] Among them, the second intermediate service recommendation information can be the result obtained by processing the second adjustment result through the second activation function of the environment joint multi-layer perceptron model.

[0122] Specifically, for each first intermediate service recommendation information, the second connection weight in the environment combined multi-layer perceptron model is used to adjust the weight, and the third connection weight adjustment result is obtained; then, the second bias of the environment combined multi-layer perceptron model is used to adjust the third connection weight adjustment result, and the second adjustment result is obtained; finally, the second adjustment result is processed using the Sigmoid activation function of the second output layer in the environment combined multi-layer perceptron model to obtain at least one second intermediate service recommendation information.

[0123] Step 606: Perform a fourth connection weight adjustment on each second intermediate service recommendation information to obtain a fourth connection weight adjustment result; adjust the fourth connection weight adjustment result to obtain a third adjustment result; according to the third adjustment result and the fourth mapping relationship, obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

[0124] Among them, the fourth connection weight adjustment result can be the result obtained by adjusting each second intermediate service recommendation information using the connection weight; among them, the fourth connection weight adjustment can be the third connection weight in the environment combined multi-layer perceptron model.

[0125] Among them, the third adjustment result can be the result obtained by passing the fourth connection weight adjustment result through the third bias of the environment combined multi-layer perceptron model.

[0126] Among them, the fourth mapping relationship can be the Sigmoid activation function of the third output layer in the environment combined multi-layer perceptron model.

[0127] Specifically, for each second intermediate service recommendation information, the third connection weight in the environment combined multi-layer perceptron model is used to adjust the weight, and the fourth connection weight adjustment result is obtained; then, the third bias of the environment combined multi-layer perceptron model is used to adjust the fourth connection weight adjustment result, and the third adjustment result is obtained; finally, the third adjustment result is processed using the Sigmoid activation function of the third output layer in the environment combined multi-layer perceptron model to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

[0128] Regarding the entire calculation process of the environment combined multi-layer perceptron model, the environment combined multi-layer perceptron model splices the business environment feature vector, user feature vector, and business feature vector, and then inputs them into the multi-layer perceptron. Through multiple neural networks, more complex and effective feature combinations are mined, so as to integrate the business environment information into the model. At the same time, the relationship between the business environment, user account, and business can be captured. The formula is as follows:

[0129]

[0130] Among them, φEMP represents the prediction result of the environmental joint multi-layer perceptron model, c m represents the business environment feature vector, p u represents the user feature vector, q i represents the business feature vector, w i and b i and f respectively represent the connection weight, bias, and activation function of the model.

[0131] In this embodiment, by refining the multiple probability iterative calculation of the feature vector splicing result by the recommendation information output layer in the environmental perception model, the calculation accuracy of the business recommendation probability of the environmental perception model can be improved, and the accuracy of the business recommendation model can be further improved.

[0132] In one embodiment, as Figure 7 shown, adjusting each first sub-business recommendation information and each second sub-business recommendation information to obtain at least one business recommendation information corresponding to the business recommendation requirement information, including:

[0133] Step 702, perform a fifth connection weight adjustment on each first sub-business recommendation information and each second sub-business recommendation information to obtain a fifth connection weight adjustment result.

[0134] Among them, the fifth connection weight adjustment result may be the result obtained by adjusting each first sub-business recommendation information and each second sub-business recommendation information using the connection weight; among them, the fifth connection weight adjustment may be one of the connection weights in the pre-trained business recommendation model.

[0135] Specifically, for each first sub-business recommendation information and each second sub-business recommendation information, use the fifth connection weight in the pre-trained business recommendation model to adjust the weight to obtain a fifth connection weight adjustment result.

[0136] Step 704, according to the fifth connection weight adjustment result and the fifth mapping relationship, obtain at least one business recommendation information corresponding to the business recommendation requirement information.

[0137] Among them, the fifth mapping relationship may be the Sigmoid activation function of one of the output layers in the pre-trained business recommendation model.

[0138] Specifically, process the fifth connection weight adjustment result using the Sigmoid activation function of the pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information.

[0139] Regarding the weight adjustment of the generalized matrix factorization model and the environment - joint multi - layer perceptron model and the processing of activation functions in the pre - trained business recommendation model. For the generalized matrix factorization model, when the business environment is relatively stable or the user account has a large amount of interaction information with the business, the model performs well; for the environment - joint multi - layer perceptron model, when the business environment changes greatly or the user account has less interaction information with the business, the model performs well. Therefore, by scoring the results of the generalized matrix factorization model and the environment - joint multi - layer perceptron model, better and more comprehensive results can be obtained. The formula is as follows:

[0140]

[0141] Among them, φ GMF and φ EMP respectively represent the output results of the generalized matrix factorization model and the environment - joint multi - layer perceptron model, and their influence on the final result is adjusted by the connection weight h. σ represents the Sigmoid activation function. Through the processing of the activation function, the result of the final multi - model prediction is output between 0 and 1. The closer to 1, the more the predicted business result of the model approaches the true needs tendency of the user account.

[0142] In this embodiment, by adjusting the weights of the calculation results of the matrix factorization model and the environment perception model, the output probability weights of the business recommendation model can be adjusted according to the stability of the business environment or the amount of interaction information between the user account and the business, making the applicable range of the business recommendation model wider.

[0143] In one embodiment, as Figure 8 shown, feature extraction is respectively performed on user account information, business information, business interaction information, and business environment information to obtain user feature vectors, business feature vectors, and business environment feature vectors corresponding to business recommendation requirements, including:

[0144] Step 802, establish an information dictionary mapping table corresponding to the business recommendation requirement information according to user account information, business information, business interaction information, and business environment information.

[0145] Among them, the information dictionary mapping table can be a table of a relationship from A to B described in the information set, and B is called the image of A under this mapping.

[0146] Specifically, for the text information in user account information, business information, business interaction information, and business environment information, an information dictionary mapping table corresponding to the business recommendation requirement information is constructed, and it is mapped and represented as a numerical value in the way of the information dictionary mapping table. Among them, the data for constructing the information dictionary mapping table are the text information of user account information, business information, and business interaction information, such as occupation, marital status, whether there is a default record, region, etc.

[0147] Step 804: Extract features from the information dictionary mapping table to obtain the user feature vector and business feature vector corresponding to the business recommendation requirement information.

[0148] Specifically, based on the information dictionary mapping table, One-hot encoding is used to extract features from each information and the mapping relationship between each information in the user account information, business information, business interaction information, and business environment information in the information dictionary mapping table, so as to obtain the user feature vector and business feature vector.

[0149] Step 806: Extract the non-text information in the user account information, business information, business interaction information, and business environment information, and construct a business environment feature vector according to the non-text information.

[0150] Specifically, for the non-text information in the user account information, business information, business interaction information, and business environment information, after feature extraction, it is represented in the way of constructing it into a multi-dimensional feature vector to obtain the business environment feature vector.

[0151] In this embodiment, by adopting different feature extraction methods for different information formats and correspondingly constructing different feature vectors, the respective characteristics of the matrix decomposition model and the environment perception model can be fully utilized to improve the accuracy of model calculation.

[0152] It should be understood that although each step in the flowcharts involved in the above embodiments is displayed in sequence according to the indication of the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, 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. These steps or stages do not necessarily need to be executed at the same moment, but can be executed at different moments. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0153] Based on the same inventive concept, an embodiment of this application further provides a business object recommendation device for implementing the business object recommendation method involved above. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the business object recommendation device provided below can refer to the limitations on a business object recommendation method in the above text, and will not be repeated here.

[0154] In one embodiment, as Figure 11 shown, a business object recommendation device is provided, including: an information acquisition module 1102, a feature extraction module 1104, an information processing module 1106, and a business recommendation module 1108, where:

[0155] The information acquisition module 1102 is configured to obtain user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information in response to the business recommendation requirement information of the resource interaction platform;

[0156] The feature extraction module 1104 is configured to respectively extract features from the user account information, business information, business interaction information, and business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement;

[0157] The information processing module 1106 is configured to input the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained by sample data corresponding to at least one business recommendation requirement information;

[0158] The business recommendation module 1108 is configured to perform business object recommendation for the user account based on at least one business recommendation information corresponding to the business recommendation requirement information.

[0159] In one embodiment, the information processing module 1106 is further configured to input the user feature vector and the business feature vector into the matrix factorization model to obtain at least one first sub-business recommendation information corresponding to the business recommendation requirement information; input the user feature vector, the business feature vector, and the business environment feature vector into the environment perception model to obtain at least one second sub-business recommendation information corresponding to the business recommendation requirement information; adjust each first sub-business recommendation information and each second sub-business recommendation information to obtain at least one business recommendation information corresponding to the business recommendation requirement information.

[0160] In one embodiment, the information processing module 1106 is further configured to multiply the user feature vector by the service feature vector to obtain a result of the dot product of the feature vectors; perform a first connection weight adjustment on the result of the dot product of the feature vectors to obtain a first connection weight adjustment result; and obtain at least one first sub-service recommendation information according to the first connection weight adjustment result and the first mapping relationship.

[0161] In one embodiment, the information processing module 1106 is further configured to splice the user feature vector, the service feature vector, and the service environment feature vector to obtain a result of the spliced feature vectors; and input the result of the spliced feature vectors into the recommendation information output layer to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

[0162] In one embodiment, the information processing module 1106 is further configured to perform a second connection weight adjustment on the result of the spliced feature vectors to obtain a second connection weight adjustment result; perform an adjustment on the second connection weight adjustment result to obtain a first adjustment result; obtain at least one first intermediate service recommendation information according to the first adjustment result and the second mapping relationship; perform a third connection weight adjustment on each first intermediate service recommendation information to obtain a third connection weight adjustment result; perform an adjustment on the third connection weight adjustment result to obtain a second adjustment result; obtain at least one second intermediate service recommendation information according to the second adjustment result and the third mapping relationship; perform a fourth connection weight adjustment on each second intermediate service recommendation information to obtain a fourth connection weight adjustment result; perform an adjustment on the fourth connection weight adjustment result to obtain a third adjustment result; and obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information according to the third adjustment result and the fourth mapping relationship.

[0163] In one embodiment, the information processing module 1106 is further configured to perform a fifth connection weight adjustment on each first sub-service recommendation information and each second sub-service recommendation information to obtain a fifth connection weight adjustment result; and obtain at least one service recommendation information corresponding to the service recommendation requirement information according to the fifth connection weight adjustment result and the fifth mapping relationship.

[0164] In one embodiment, the feature extraction module 1104 is further configured to establish an information dictionary mapping table corresponding to the service recommendation requirement information according to the user account information, the service information, the service interaction information, and the service environment information; perform feature extraction on the information dictionary mapping table to obtain the user feature vector and the service feature vector corresponding to the service recommendation requirement information; and extract the non-text information from the user account information, the service information, the service interaction information, and the service environment information, and construct a service environment feature vector according to the non-text information.

[0165] Each module in the above-mentioned business object recommendation device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0166] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 12 shown. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store server data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a business object recommendation method.

[0167] Those skilled in the art can understand that Figure 12 the structure shown in

[0168] is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0169] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the steps in the above method embodiments are implemented.

[0170] In one embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0171] 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 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.

[0172] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. 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 methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0173] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered to be within the scope described in this specification.

[0174] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A business object recommendation method, characterized in that, The method includes: In response to the business recommendation requirement information of the resource interaction platform, obtaining the user account information, business information, business interaction information, and business environment information corresponding to the business recommendation requirement information; Respectively performing feature extraction on the user account information, the business information, the business interaction information, and the business environment information to obtain a user feature vector, a business feature vector, and a business environment feature vector corresponding to the business recommendation requirement; Inputting the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information; the business recommendation information represents the matching degree between the business and the user account; the pre-trained business recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained through sample data corresponding to at least one of the business recommendation requirement information; Based on at least one business recommendation information corresponding to the business recommendation requirement information, performing business object recommendation for the user account; The inputting the user feature vector, the business feature vector, and the business environment feature vector into a pre-trained business recommendation model to obtain at least one business recommendation information corresponding to the business recommendation requirement information includes: Inputting the user feature vector and the business feature vector into a matrix factorization model to obtain at least one first sub-business recommendation information corresponding to the business recommendation requirement information; Inputting the user feature vector, the business feature vector, and the business environment feature vector into an environment perception model to obtain at least one second sub-business recommendation information corresponding to the business recommendation requirement information; Adjusting each of the first sub-business recommendation information and each of the second sub-business recommendation information to obtain at least one business recommendation information corresponding to the business recommendation requirement information; The adjusting each of the first sub-business recommendation information and each of the second sub-business recommendation information to obtain at least one business recommendation information corresponding to the business recommendation requirement information includes: Performing fifth connection weight adjustment on each of the first sub-business recommendation information and each of the second sub-business recommendation information to obtain a fifth connection weight adjustment result; According to the fifth connection weight adjustment result and a fifth mapping relationship, obtaining at least one business recommendation information corresponding to the business recommendation requirement information.

2. The method according to claim 1, wherein The inputting the user feature vector and the business feature vector into a matrix factorization model to obtain at least one first sub-business recommendation information corresponding to the business recommendation requirement information includes: Multiplying the user feature vector by the business feature vector to obtain a feature vector multiplication result; Performing first connection weight adjustment on the feature vector multiplication result to obtain a first connection weight adjustment result; According to the first connection weight adjustment result and a first mapping relationship, obtaining at least one first sub-business recommendation information.

3. The method according to claim 1, characterized in that, Inputting the user feature vector, the service feature vector, and the service environment feature vector into an environment perception model to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information includes: Concatenating the user feature vector, the service feature vector, and the service environment feature vector to obtain a concatenated feature vector result; Inputting the concatenated feature vector result into a recommendation information output layer to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information.

4. The method according to claim 3, wherein The step of inputting the concatenated feature vector result into a recommendation information output layer to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information includes: Performing a second connection weight adjustment on the concatenated feature vector result to obtain a second connection weight adjustment result; adjusting the second connection weight adjustment result to obtain a first adjustment result; obtaining at least one first intermediate service recommendation information according to the first adjustment result and a second mapping relationship; Performing a third connection weight adjustment on each of the first intermediate service recommendation information to obtain a third connection weight adjustment result; adjusting the third connection weight adjustment result to obtain a second adjustment result; obtaining at least one second intermediate service recommendation information according to the second adjustment result and a third mapping relationship; Performing a fourth connection weight adjustment on each of the second intermediate service recommendation information to obtain a fourth connection weight adjustment result; adjusting the fourth connection weight adjustment result to obtain a third adjustment result; obtaining at least one second sub-service recommendation information corresponding to the service recommendation requirement information according to the third adjustment result and a fourth mapping relationship.

5. The method according to claim 1, characterized in that The step of respectively performing feature extraction on the user account information, the service information, the service interaction information, and the service environment information to obtain a user feature vector, a service feature vector, and a service environment feature vector corresponding to the service recommendation requirement includes: Establishing an information dictionary mapping table corresponding to the service recommendation requirement information according to the user account information, the service information, the service interaction information, and the service environment information; Performing feature extraction on the information dictionary mapping table to obtain a user feature vector and a service feature vector corresponding to the service recommendation requirement information; Extracting non-text information from the user account information, the service information, the service interaction information, and the service environment information, and constructing the service environment feature vector according to the non-text information.

6. A business object recommendation device, characterized in that, The apparatus includes: An information acquisition module, configured to acquire user account information, service information, service interaction information, and service environment information corresponding to the service recommendation requirement information in response to the service recommendation requirement information of the resource interaction platform; A feature extraction module, configured to respectively perform feature extraction on the user account information, the service information, the service interaction information, and the service environment information to obtain a user feature vector, a service feature vector, and a service environment feature vector corresponding to the service recommendation requirement; An information processing module, configured to input the user feature vector, the service feature vector, and the service environment feature vector into a pre-trained service recommendation model to obtain at least one service recommendation information corresponding to the service recommendation requirement information; the service recommendation information represents the matching degree between the service and the user account; the pre-trained service recommendation model includes a matrix factorization model and an environment perception model; the matrix factorization model and the environment perception model are trained by sample data corresponding to at least one of the service recommendation requirement information; A service recommendation module, configured to perform service object recommendation for the user account based on at least one service recommendation information corresponding to the service recommendation requirement information; The information processing module is further configured to input the user feature vector and the service feature vector into the matrix factorization model to obtain at least one first sub-service recommendation information corresponding to the service recommendation requirement information; input the user feature vector, the service feature vector, and the service environment feature vector into the environment perception model to obtain at least one second sub-service recommendation information corresponding to the service recommendation requirement information; adjust each of the first sub-service recommendation information and each of the second sub-service recommendation information to obtain at least one service recommendation information corresponding to the service recommendation requirement information; The information processing module is further configured to perform fifth connection weight adjustment on each of the first sub-service recommendation information and each of the second sub-service recommendation information to obtain a fifth connection weight adjustment result; and obtain at least one service recommendation information corresponding to the service recommendation requirement information according to the fifth connection weight adjustment result and the fifth mapping relationship.

7. A computer device, comprising a memory and a processor, the memory storing 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 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images