Recommendation Method, Device, Equipment, Medium and Program Product for Automobile Products

Abstract

The present invention provides a recommendation method, device, equipment, medium and program product for automotive products, which relates to the field of artificial intelligence and can improve the prediction accuracy and recommendation effectiveness in scenarios where information related to purchasing automobiles is relatively sparse. The recommendation method for automotive products includes: obtaining the data of the first user in the banking system with the authorization of the first user to obtain the first initial data, then screening out M feature data most relevant to the user's car purchase intention from it to obtain the first feature set, and constructing a first prompt instruction represented in natural language by using the first feature set; inputting the first prompt instruction into the user car purchase intention recognition model, and obtaining the first car purchase intention recognition result output by the user car purchase intention recognition model, where the user car purchase intention recognition model is obtained by fine-tuning a pre-trained large language model; when the first car purchase intention recognition result meets the potential car purchase condition, recommending automotive products to the first user.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and can be used in the field of fintech or other fields. More specifically, it relates to a method, device, equipment, medium and program product for recommending automotive products. Background Art

[0002] With the popularization of automobile consumption and the improvement of economic level, automobile installment products have become the choice of more and more consumers when purchasing cars. At the same time, with the development of artificial intelligence, the prediction and recommendation of automotive products through machine learning models have also been adopted by more and more enterprises.

[0003] In the process of implementing the inventive concept of the present invention, the inventor found that there are the following defects in predicting automotive products through machine learning models in the related art: The related art usually uses user car purchase behavior data (such as user browsing of vehicle models, inquiry, etc.) to train a model to predict the automotive products that a user may purchase, and it is unable to effectively predict users who have not shown obvious car purchase behavior or lack relevant behavior data. In fact, the number of users who have not purchased a car or have not shown obvious car purchase behavior or lack relevant behavior data is much larger than the amount of data of users who have purchased a car or have car purchase behavior. That is to say, when users who have not purchased a car or have not shown obvious car purchase behavior or lack relevant behavior data are also included in the prediction scope, the information related to car purchase that the model can learn will become very sparse, resulting in inaccurate model prediction. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method and device for recommending automotive products that can make effective recommendations in scenarios where information related to car purchase is relatively sparse, a method and device for training a user car purchase willingness recognition model based on a large language model, as well as an electronic device, a medium and a program product.

[0005] In the first aspect of the embodiments of the present invention, a method for recommending automotive products is provided. The method includes: obtaining authorization from a first user to use their data in the banking system; after obtaining the authorization from the first user, obtaining the data of the first user in the banking system to obtain first initial data; screening out M feature data corresponding to M features from the first initial data to obtain a first feature set, where the M features are a set of features most relevant to the user's car purchase intention screened out by a feature engineering algorithm, and M is an integer greater than or equal to 2; constructing a first prompt instruction represented in natural language using the first feature set; inputting the first prompt instruction into a user car purchase intention recognition model and obtaining a first car purchase intention recognition result output by the user car purchase intention recognition model, where the user car purchase intention recognition model is obtained by fine-tuning a pre-trained large language model; and when the first car purchase intention recognition result meets the potential car purchase condition, recommending automotive products to the first user.

[0006] In the second aspect of the embodiments of the present invention, a training method for a user car purchase intention recognition model based on a large language model is provided. The method includes: obtaining authorization from a second user to use their data in the banking system, where the second user is a user who has purchased an automotive installment product in the banking system or a user who has not purchased any automotive-related products in the banking system; after obtaining the authorization from the second user, obtaining the data of the second user in the banking system to obtain second initial data; screening out M feature data corresponding to M features from the second initial data to obtain a second feature set, where the M features are a set of features most relevant to the user's car purchase intention screened out by a feature engineering algorithm, and M is an integer greater than or equal to 2; when the second user is the user who has purchased a car, labeling the second feature set with the information of having purchased a car to obtain a positive sample data; and when the second user is the user who has not purchased a car, labeling the second feature set with the information of not having purchased a car to obtain a negative sample data; after obtaining an equal amount of the positive sample data and the negative sample data, for each piece of the positive sample data and each piece of the negative sample data, respectively constructing a prompt instruction represented in natural language to obtain an instruction fine-tuning data set; and using the instruction fine-tuning data set to fine-tune the large language model to obtain the user car purchase intention recognition model.

[0007] In the third aspect of the embodiments of the present invention, a recommendation device for automotive products is provided. The device includes: a first acquisition module, a feature engineering and feature analysis module, a purchase intention prediction module, and an automotive product recommendation module. The first acquisition module is configured to obtain the authorization of the first user to use their data in the banking system, and after obtaining the authorization of the first user, obtain the data of the first user in the banking system to obtain first initial data. The feature engineering and feature analysis module is configured to screen out M feature data corresponding to M features from the first initial data to obtain a first feature set, where the M features are a set of features most relevant to the user's purchase intention screened out by a feature engineering algorithm, and M is an integer greater than or equal to 2. The purchase intention prediction module is configured to construct a first prompt instruction represented in natural language using the first feature set, and input the first prompt instruction into the user purchase intention recognition model, and obtain a first purchase intention recognition result output by the user purchase intention recognition model, where the user purchase intention recognition model is obtained by fine-tuning and training a pre-trained large language model. The automotive product recommendation module is configured to recommend automotive products to the first user when the first purchase intention recognition result meets the potential purchase condition.

[0008] In the fourth aspect of the embodiments of the present invention, a training device for a user purchase intention recognition model based on a large language model is provided. The training device includes a second acquisition module, a feature engineering and feature analysis module, and a fine-tuning training module. The second acquisition module is configured to obtain the authorization of the second user to use their data in the banking system, and after obtaining the authorization of the second user, obtain the data of the second user in the banking system to obtain second initial data, where the second user is a user who has purchased an automotive installment product in the banking system or a user who has not purchased any automotive-related products in the banking system. The feature engineering and feature analysis module is configured to: screen out M feature data corresponding to M features from the second initial data to obtain a second feature set, where the M features are a set of features most relevant to the user's purchase intention screened out by a feature engineering algorithm, and M is an integer greater than or equal to 2; when the second user is the user who has purchased the automotive installment product, label the second feature set with the information of having purchased a car to obtain a positive sample data; and when the second user is the user who has not purchased any automotive-related products, label the second feature set with the information of not having purchased a car to obtain a negative sample data. The fine-tuning training module is configured to, after obtaining an equal amount of the positive sample data and the negative sample data, respectively construct a prompt instruction represented in natural language for each piece of the positive sample data and each piece of the negative sample data to obtain an instruction fine-tuning data set, and use the instruction fine-tuning data set to perform fine-tuning training on the large language model to obtain the user purchase intention recognition model.

[0009] In the fifth aspect of the embodiments of the present invention, an electronic device is provided. The electronic device includes: one or more processors and a memory. The memory is used to store one or more computer programs. The one or more processors execute the one or more computer programs to implement the steps of the method provided in the first aspect or the second aspect above.

[0010] In the sixth aspect of the embodiments of the present invention, a computer-readable storage medium stores a computer program or instruction, and when the computer program or instruction is executed by a processor, the steps of the method provided in the first aspect or the second aspect above are implemented.

[0011] In the seventh aspect of the embodiments of the present invention, a computer program product includes a computer program or instruction, and when the computer program or instruction is executed by a processor, the steps of the method provided in the first aspect or the second aspect above are implemented.

[0012] The above one or more embodiments have the following advantages or beneficial effects: First, important features are screened out from the initial data, and then a user car purchase intention recognition model constructed based on a large language model is used to identify potential car purchase users. With the feature extraction of the feature engineering algorithm and the general understanding ability of the large language model, it can effectively solve the problem of sparse car purchase-related data caused by the fact that the number of users who have not purchased a car or have not shown obvious car purchase behaviors or lack relevant behavior data is much larger than the data volume of users who have purchased a car or have car purchase behaviors. Therefore, the understanding and prediction ability of the user's car purchase intention can be improved, the pertinence of car product recommendations can be enhanced, and in the embodiments of the present invention, recommendations are made to users only when the user car purchase intention recognition result meets the potential car purchase conditions, so as to reduce unnecessary disturbances to users while expanding the scope of the predicted population, improve the user experience and purchase conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Through the following description of the embodiments of the present invention with reference to the drawings, the above content and other objects, features, and advantages of the present invention will become clearer. In the drawings:

[0014] Figure 1 Schematically shows the conceptual framework of a method and device for recommending car products according to an embodiment of the present invention;

[0015] Figure 2 Schematically shows the application scenarios of a method, device, equipment, medium, and program product for recommending car products according to an embodiment of the present invention;

[0016] Figure 3 Schematically shows the flowchart of a method for recommending car products according to an embodiment of the present invention;

[0017] Figure 4The flowchart of the method for training a user's car purchase intention recognition model according to an embodiment of the present invention is schematically shown, wherein the user's car purchase intention recognition model is obtained by fine-tuning and training a pre-trained large language model;

[0018] Figure 5 The overall flowchart of each model in the training stage according to an embodiment of the present invention is schematically shown;

[0019] Figure 6 The flowchart of the method for recommending car products according to another embodiment of the present invention is schematically shown;

[0020] Figure 7 The block diagram of the device for recommending car products according to an embodiment of the present invention is schematically shown;

[0021] Figure 8 The block diagram of the device for training the user's car purchase intention recognition model according to an embodiment of the present invention is schematically shown; and

[0022] Figure 9 The block diagram of the electronic device according to an embodiment of the present invention is schematically shown. Detailed implementation manners

[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, many specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present invention. However, it is obvious that one or more embodiments can be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0024] In the technical solution of the present invention, the user information involved (including but not limited to data in the bank system, user personal information, signing records, user image information, user device information, such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant data and other processing all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0025] In the scenario of making automated decisions using personal information, the method, device, and system provided by the embodiments of the present invention all provide corresponding operation entrances for users to choose to agree or reject the automated decision results; if the user chooses to reject, the expert decision-making process will be entered. The expression "automated decision" here refers to the activity of automatically analyzing and evaluating an individual's behavior habits, hobbies, or economic, health, credit status, etc. through a computer program and making a decision. The expression "expert decision" here refers to the activity of making a decision by a person who specializes in a certain field, has specialized experience, knowledge, and skills, and has reached a certain professional level.

[0026] The embodiments of the present invention provide a method, device, equipment, medium, and program product for recommending automotive products. When obtaining the authorization of the user to use their data in the banking system, the user's car purchase intention can be analyzed based on the user's data in the banking system, and when it is determined that the user has a relatively high car purchase intention, automotive products can be recommended to the user specifically. Among them, Figure 1 Schematically shows the conceptual framework of the method and device for recommending automotive products according to the embodiments of the present invention.

[0027] Refer to Figure 1 , considering that the data of the user in the banking system may lack data related to car purchase behaviors (such as browsing car models, making inquiries), and there may also be feature redundancy, which interferes with the accuracy of prediction. The embodiments of the present invention will first screen out the feature data with relatively high correlation with the car purchase intention from the obtained user data in the banking system through feature engineering, reducing the interference of factors such as feature redundancy on the prediction result when directly using the user data in the banking system.

[0028] Furthermore, considering that the information related to the user's car purchase in the banking system mainly comes from the signing records of car installment products, and thus the information related to the user's car purchase in the banking system is significantly smaller than the scale and data volume of car purchase users in the entire market, resulting in the problem of sparse data volume that the model can learn. The embodiments of the present invention use a user car purchase intention recognition model obtained by fine-tuning and training a pre-trained large language model to recognize the user's car purchase intention. In this way, the ability of the large language model to convert prompt instructions expressed in natural language into high-dimensional semantic representations can be utilized to further improve the expression ability of features, capture the context and emotional information therein, thereby improving the understanding and prediction ability of the user's car purchase intention, and more accurately identifying potential car purchase users.

[0029] It can be seen that after screening out important features, the embodiments of the present invention can effectively solve problems such as feature extraction and combination, non-linear relationship modeling, and sparsity of purchase-related data in the banking system when using the user purchase intention recognition model constructed based on the large language model to identify potential purchase customer groups, improve the understanding and prediction ability of the user's purchase intention, thereby enhancing the pertinence of automobile product recommendations, and improving the user experience and purchase conversion rate.

[0030] Figure 2 Schematically shows an application scenario of a method, apparatus, device, medium, and program product for recommending automobile products according to an embodiment of the present invention.

[0031] As Figure 2 shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0032] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications, such as mobile banking clients, may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103.

[0033] The server 105 may be a server that provides various services, such as a background management server (only for example) that supports the websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103.

[0034] It should be noted that the method for recommending automobile products and / or the training method of the user purchase intention recognition model provided by the embodiments of the present invention can generally be executed by the server 105. Correspondingly, the apparatus for recommending automobile products and / or the training apparatus of the user purchase intention recognition model provided by the embodiments of the present invention can generally be set in the server 105.

[0035] It should be understood that Figure 2 the numbers of terminal devices, networks, and servers in

[0036] are merely illustrative. According to actual needs, there can be any number of terminal devices, networks, and servers. Figure 2 The following will describe in detail the method and apparatus for automobile products according to the embodiments of the present invention based on the

[0037] Figure 3 The flowchart of the recommendation method for automotive products according to an embodiment of the present invention is schematically shown.

[0038] As Figure 3 shown, the recommendation method for automotive products in this embodiment may include operation S301 to operation S306, and this method may be executed by the server 105.

[0039] In operation S301, obtain the authorization of the first user to use their data in the banking system. For example, the server 105 requests the user's authorization through interaction with the first terminal device 101, the second terminal device 102, or the third terminal device 103.

[0040] In operation S302, after obtaining the authorization of the first user, obtain the data of the first user in the banking system to obtain the first initial data. Among them, the data of the user in the banking system may be, for example, user basic information, transaction records, and other data.

[0041] In operation S303, screen out M feature data corresponding to M features from the first initial data to obtain the first feature set. The M features are a set of features most relevant to the user's car purchase intention screened out by the feature engineering algorithm, where M is an integer greater than or equal to 2.

[0042] The process of screening M features by the feature engineering algorithm has been completed before operation S301. In one embodiment, through the feature engineering algorithm, after obtaining the corresponding user authorization, based on a large amount of user data accumulated in the banking system, combined with information such as whether the user purchases automotive installment products, a corresponding model can be established to screen out important features related to the purchase intention of automotive installment products.

[0043] The methods for screening features by the feature engineering algorithm include, but are not limited to: feature importance based on tree models, feature importance based on linear models, feature importance based on feature selection, feature importance based on neural networks, and Shapley Additive exPlanations (SHAP) values, etc.

[0044] In one embodiment, the feature engineering algorithm is specifically constructed using a random forest model.

[0045] Accordingly, the screening process of the M features can be as follows. First, after obtaining the user's authorization to use their data in the banking system, a random forest model is trained to predict the user's willingness to purchase a car based on the user's data in the banking system. For example, a training set composed of the data of an equal number of users who have purchased cars (users who have purchased car installment products) and users who have not purchased cars (users who have not purchased car installment products) in the banking system is used to train the random forest model. Among them, corresponding labels are assigned to the users who have purchased cars and the users who have not purchased cars respectively to train the random forest model to predict the probability of a user purchasing a car. Then, after the random forest model is trained, based on the importance of the input features of the random forest model and the Shapley Additive Explanation (SHAP) values, M features are screened out. In this way, through the feature importance and SHAP values in the random forest model, a series of user features related to the user's willingness to purchase a car are screened out, improving the feature pertinence of the subsequent large language model processing and reducing the interference of redundant features on the large language model analysis.

[0046] Furthermore, in another embodiment, the feature engineering algorithm can be implemented through multiple random forest models, and each random forest model learns a car purchase attribute (such as price, category, or energy type).

[0047] Accordingly, the screening process of the M features can be as follows. First, obtain the authorization of at least one user who has purchased a car to use their data in the banking system, where the user who has purchased a car is a user who has purchased a car installment product in the banking system. Then, after obtaining the authorization of the user who has purchased a car, obtain the data of the user who has purchased a car in the banking system, and obtain n labels of the user who has purchased a car based on the category information of n attributes of the car in the car installment product purchased by the user who has purchased a car, where n is an integer greater than or equal to 2. Next, train n random forest models, each random forest model is used to predict one of the n attributes. Among them, when training each random forest model, the user who has purchased a car is labeled with the label corresponding to the attribute to be predicted by the random forest model, and the data of the user who has purchased a car in the banking system is used as the input to train the random forest model. Finally, after the n random forest models are trained, based on the importance of the input features of the n random forest models and the Shapley Additive Explanation (SHAP) values, M features are screened out.

[0048] Screen features through multiple random forest models. The advantage of each random forest model learning a car purchase attribute is that, on the one hand, it can construct a random forest model only using the data of users who have purchased cars in the bank system, avoiding the problem of incomplete and inaccurate information that may exist when using the data of users who have not purchased cars in the bank system, because there are cases where users purchase cars through other channels rather than subscribing to bank car installment products; on the other hand, since the data in the bank system is mainly the financial data of users who purchase installment products and not from car purchase behavior data, and there is a problem of data sparsity, modeling and feature screening for different attributes of car products respectively can improve the pertinence of modeling, enhance the accuracy and comprehensiveness of feature screening, increase the number of learning times and dimensions of the model, and alleviate or even overcome the defect of data sparsity to a certain extent.

[0049] Next, in operation S304, construct a first prompt instruction represented in natural language using the first feature set. The prompt instruction can be constructed through a specially designed instruction framework. For example, the following prompt instruction can be constructed:

[0050] "Please assess the likelihood of this individual purchasing a car in the near future, presented on a scale of 0-1 and briefly explain the reason of this score: Gender: female; Age: 37; Education: Bachelor's degree..."

[0051] Among them, Gender, Education, Bachelor's degree, etc. are the M features selected, and the order of these features in the prompt instruction can be random.

[0052] Then, in operation S305, input the first prompt instruction into the user car purchase intention recognition model, and obtain the first car purchase intention recognition result output by the user car purchase intention recognition model. Among them, the user car purchase intention recognition model is obtained by fine-tuning and training a pre-trained large language model. The first car purchase intention recognition result can be represented by, but not limited to, a score value, a probability value, a classification result, etc.

[0053] Finally, in operation S306, when the first car purchase intention recognition result meets the potential car purchase condition, recommend car products to the first user. The potential car purchase condition can be, for example, that the score or probability value in the car purchase intention recognition result exceeds a threshold, or ranks top N in a large user group, or is classified into a high car purchase intention category, etc.

[0054] In one embodiment, accurate recommendations can be made by constructing an automotive product recommendation model. Specifically, the first initial data (which can also be the first feature set in some embodiments) can be input into the automotive product recommendation model, and the recommended category information of the automotive products output by the automotive product recommendation model can be obtained. Then, the automotive products indicated by the recommended category information are recommended to the first user. The recommended category information can include, for example, the category information of multiple attributes (such as price, energy type, category, etc.). In this way, for users who meet the potential car purchase conditions, automotive products of types with relatively high purchase likelihood can be specifically recommended, improving the pertinence and conversion rate of automotive product recommendations and enhancing the user experience. It also improves the effectiveness of the marketing and promotion of bank automotive installment products.

[0055] For the automotive product recommendation method of the embodiments of the present invention, when making automotive product recommendations, the feature extraction of the feature engineering algorithm and the general understanding ability of the large language model can be utilized to solve technical problems such as feature extraction and combination, non-linear relationship modeling, and data sparsity faced in the process of identifying potential car purchase customer groups based on the data in the bank system, improving the understanding and prediction ability of users' car purchase intentions, thereby enhancing the user experience and purchase conversion rate.

[0056] The user car purchase intention recognition model used in the above operation S305 is obtained by fine-tuning and training a pre-trained large language model during the training stage before operation S301. Among them, Figure 4 The flowchart of the training method of the user car purchase intention recognition model according to an embodiment of the present invention is schematically shown.

[0057] Refer to Figure 4 , according to this embodiment, the process of fine-tuning and training the large language model to obtain the user car purchase intention recognition model can include operation S401 to operation S406.

[0058] First, in operation S401, authorization from a second user to use the data in their bank system is obtained. The second user is a user who has purchased an automotive installment product in the bank system or a non-car-purchasing user who has not used any automotive-related products in the bank system.

[0059] When selecting non-car-purchasing users, various features related to car purchase are excluded as much as possible. For example, it is not only required that the user has not purchased an automotive installment product in the bank system, but also users with automotive maintenance consumption transactions and automotive fuel consumption transactions are excluded through the bank transaction process, so as to ensure that the selection of non-car-purchasing users is as objective and accurate as possible.

[0060] In operation S402, after obtaining the authorization of the second user, the data of the second user in the bank system is obtained to obtain the second initial data.

[0061] In operation S403, M feature data corresponding to M features are screened out from the second initial data to obtain a second feature set, where the M features are a set of features most relevant to the user's car purchase intention screened out by the above-mentioned feature engineering algorithm.

[0062] In operation S404, when the second user is a user who has purchased a car, the information of having purchased a car is used as a label to annotate the second feature set to obtain a positive sample data; and when the second user is a user who has not purchased a car, the information of not having purchased a car is used as a label to annotate the second feature set to obtain a negative sample data.

[0063] In operation S405, after obtaining an equal amount of positive sample data and negative sample data, for each positive sample data and each negative sample data, a prompt instruction represented in natural language is constructed respectively to obtain an instruction fine-tuning data set. Among them, when fine-tuning and training the large language model and when using the user car purchase intention recognition model in the application stage, the instruction frameworks for constructing the fine-tuning instructions are basically the same.

[0064] In operation S406, the large language model is fine-tuned and trained using the instruction fine-tuning data set to obtain a user car purchase intention recognition model.

[0065] Through the fine-tuning and training of the pre-trained large language model in the embodiments of the present invention, the large language model can fully learn the relationship between each feature and the car purchase intention from the important features related to the car purchase intention of users who have purchased cars and users who have not purchased cars, which can be specifically manifested in three aspects: effective extraction and combination of semantic features, learning and modeling of the non-linear relationship between features and the car purchase intention, and effectively coping with data sparsity:

[0066] (1) Make full use of the general understanding ability of the large language model to effectively extract and combine semantic features. Among them, the large language model can convert the text information expressed in natural language in the instruction fine-tuning data set into a high-dimensional semantic representation, capturing the context and emotional information therein. This representation can be used as rich features for predicting the car purchase intention. In addition, the large language model can also automatically learn the correlation and interaction relationship between features, further improving the expression ability of features.

[0067] (2) Learning and modeling of the non-linear relationship between features and the car purchase intention: The formation of the car purchase intention is affected by multiple factors, and the complex non-linear relationship often cannot be accurately modeled by traditional linear models. Through the deep neural network structure, the large language model can layer by layer extract the high-level abstract representation of the data, so as to better understand the complex relationship in the data; through the non-linear transformation of the activation function and the weight matrix, the large language model can learn the non-linear patterns in the data, improving the prediction accuracy of the car purchase intention.

[0068] (3) It can make accurate predictions in scenarios of data sparsity: As introduced above, when predicting the willingness to purchase a car based on the data in the banking system, the input data is relatively sparse. Data such as users' behavioral data and preference information is either scarce or missing. Traditional models may perform poorly when dealing with sparse data, while large language models have certain advantages in this regard. By fine-tuning a pre-trained large language model and training with a large amount of data, the large language model can learn the distribution characteristics and potential structures in the data, improve the modeling ability for sparse data, and thus better utilize the sparse data to explore the deep connection between features and the willingness to purchase a car. In addition, the large language model can also focus on important features through techniques such as the attention mechanism to further improve the accuracy of prediction.

[0069] It can be seen that the embodiment of the present invention obtains a user car purchase willingness recognition model through fine-tuning and training of a pre-trained large language model, which can effectively solve technical problems such as semantic feature extraction and combination, non-linear relationship modeling, and data sparsity processing, and improve the understanding and prediction ability of users' car purchase willingness based on the data in the banking system. Thus, it can effectively mine the car purchase behaviors of users who have not shown obvious car purchase behaviors or lack relevant data, so that the prediction of car products is not limited to the user group that can obtain relevant car purchase behavior data, expanding the scope of the predicted population for purchasing car products and increasing the purchase conversion rate of car products.

[0070] The embodiment of the present invention can effectively overcome problems such as that many data when predicting car products using the data in the banking system are not from car purchase behaviors (such as browsing car models, making inquiries), and the information related to users' car purchases in the banking system is relatively sparse (such as the proportion of users purchasing car installment products is small and the information is incomplete, etc.). Therefore, it can effectively improve the mining of the potential car purchase behaviors of users who have not purchased a car or have not shown obvious car purchase behaviors or lack relevant behavior data, and improve the identification and recommendation ability for potential car purchase users.

[0071] The following will further use Figure 5 and Figure 6 to elaborate in detail on the overall process framework of the model training stage and the application stage of the car product recommendation method according to an embodiment of the present invention. It can be understood that the following description is only exemplary and does not constitute a limitation to the present invention.

[0072] Figure 5 Schematically shows the overall flowchart of each model in the training stage in an embodiment of the present invention, describing the overall calculation framework of the training stage.

[0073] In this embodiment, the models to be constructed in the training phase include: a random forest model, a user car purchase intention recognition model, and a car product recommendation model. Among them, the training of the random forest model is used to construct a feature engineering and feature analysis module, the purpose of which is to screen out the M features most relevant to the user's car purchase intention from the initial data of the user in the bank system according to the random forest model. The user car purchase intention recognition model is obtained by fine-tuning the pre-trained large language model and is used to recognize the user's car purchase intention. The car product recommendation model is used to predict the type information of the car products that the user may purchase.

[0074] In this embodiment, first, the feature engineering and feature analysis module is used to obtain the highly important features most relevant to the user's car purchase intention from the input existing user car analysis records, then these highly important features are used to construct prompt instructions to obtain a fine-tuning dataset, and then these fine-tuning datasets are used to fine-tune the pre-trained large language model to obtain the user car purchase intention recognition model. Finally, the relevant data in the input existing user car analysis records are used to train the car product recommendation model.

[0075] First, the feature engineering and feature analysis module is constructed through the following steps (1) and (2).

[0076] Step (1), after obtaining the user's authorization to use their data, obtain the signing records of users who have signed car installment products.

[0077] Suppose a series of records obtained are in the form of where represents the user, and represents the car purchased by the user. By associating the personal basic information registered by each user in the bank system and the relevant features of the transaction-generated flow, a complete car installment dataset is obtained: where is the feature of the j-th user, including user basic features, financial features, and non-financial features. In this embodiment, each car model can have features of three attributes: price, energy type, and category.

[0078] Step (2), construct a random forest model for each of the above three attributes, so a total of three random forest models are constructed. All three random forest models take all the features included in the complete car installment dataset sorted out in step (1) as input, and use the complete car installment dataset in step (1) for supervised training with the price range, energy type, and car category of the cars purchased by the user as the prediction targets respectively.

[0079] In one embodiment, the value ranges of the automobile price range, energy type, and automobile category can be set as follows. The automobile price range can be divided into 11 categories: "0 - 50,000", "50,000 - 80,000", "80,000 - 120,000", "120,000 to 150,000", "150,000 - 180,000", "180,000 - 250,000", "250,000 - 350,000", "350,000 - 500,000", "500,000 - 800,000", "800,000 - 1,000,000", and "above 1,000,000". The automobile energy type can be divided into 2 categories: "conventional energy vehicle" and "new energy vehicle". The automobile category can be divided into 6 categories: "mid - size vehicle", "compact vehicle", "compact SUV", "mid - size SUV", "large - mid - size vehicle", and "others".

[0080] After the three random forest models are trained, features are screened based on the importance of the input features and the SHAP (Shapley Additive Explanation) values of each random forest model.

[0081] Among them, the feature importance can be calculated based on the method of mean decrease in impurity. The formula for mean decrease in impurity is:

[0082]

[0083] Where is the feature to be measured, including: basic features such as user age and gender, financial features such as transaction flow, and non - financial features such as user education level and marital status.

[0084] is the number of trees in the random forest model, is the Gini impurity of the node before splitting in the th tree. is the weighted sum (weighted by the proportion of child nodes) of the Gini impurities of the two child nodes after splitting of the feature in the th tree. Therefore,

[0085] represents the value of the decrease in Gini impurity after node splitting. The higher the sum of the values of the decrease in Gini impurity of all nodes, the greater the influence of the feature on the decision. The calculation method of the SHAP value is as follows: First, a part of the samples are randomly selected from the complete automobile installment dataset for the calculation of the SHAP value (for example, 10,000 samples are used). Subsequently, for each sample, each feature in the decision tree is traversed, and the contribution of each feature to the prediction of a specific sample is evaluated by considering all possible feature subsets. Finally, the path of the feature in the decision tree, that is, the path from the root node to the leaf node, which determines how the feature affects the final prediction, is calculated. After the calculation of each sample is completed, the SHAP values of each tree in the random forest are averaged to obtain the overall SHAP value.

[0086] Based on the feature importance and the absolute value of SHAP values, the most relevant features (such as ) are selected as the input features for fine-tuning the large language model. For example, the weighted values of feature importance and SHAP values can be set as the basis for screening. Among them, when screening from three random forest models, it can be to separately screen out the same number of features from each random forest model, and then summarize and remove duplicates; or it can also be to combine the decision trees in the three random forest models and screen out M features from them. The specific screening method is not limited in the present invention.

[0087] Next, the user car purchase intention recognition model based on the large language model is constructed through the following steps (3) to (6).

[0088] Step (3), first among the users in the bank system, after obtaining the authorization of the users to use their data, randomly select the same number of users who have not signed the car installment product (users who have not purchased a car) as the number of users who signed the car installment product in step (1). When selecting users who have not signed the car installment product (users who have not purchased a car), try to fully exclude users who contain other features related to car purchase (such as car maintenance consumption, car fuel consumption), so as to ensure that among the selected users who have not signed the car installment product, as few users who have actually purchased a car as possible are included.

[0089] Then, the initial data of the users who signed the car installment product obtained in step (1) and the newly obtained users who have not signed the car installment product in the bank system are subjected to feature extraction or screening according to the features screened out in step (2) to obtain the labeled data set as follows:

[0090]

[0091] Among them, is the randomly selected user who has not signed the car installment product, are the important features screened out in step (2), and "1" in the last dimension indicates having a car purchase record, and "0" indicates not having a car purchase record.

[0092] Step (4), use the labeled data set obtained in step (3) to construct a fine-tuning data set. An example of the construction method is as follows.

[0093] Suppose the sample corresponds to the user features (gender: male; age: 27; education level: undergraduate;...), then the prompt instruction expressed in natural language can be: = "Please estimatethe likelihood of this individual purchasing a car, with a value ranging from0 to 1: Gender: Male; Age: 27; Education: Bachelor's degree...", and its corresponding instruction fine-tuning sample is: ( , 1), which is a positive sample data. It should be noted that when constructing the instruction fine-tuning sample, the order in which each feature appears in the prompt instruction can be random to improve the robustness of the fine-tuned large language model.

[0094] All samples of the labeled dataset can be constructed in a similar way to obtain the following instruction fine-tuning dataset:

[0095] .

[0096] In step (5), use the instruction fine-tuning dataset obtained in step (4) to perform instruction fine-tuning on the large language model. Specifically, the Low-Rank Adaptation (LoRA) method for fine-tuning the large language model based on the low-rank matrix factorization technique can be used to perform instruction fine-tuning on the large language model.

[0097] The core idea of LoRA is to introduce two low-rank matrices into the weight matrix of the pre-trained large language model to achieve fine-tuning of the model. For each weight matrix in the large language model , LoRA is updated in the following way:

[0098]

[0099] Among them, and are two low-rank matrices, and their product is used to adjust the original weight matrix . Since and have a much lower rank than , this method only needs to introduce a small number of additional parameters, thus reducing the storage requirements and computational costs of the model.

[0100] During fine-tuning training, first load the pre-trained large language model, including its weights and configuration, and then initialize two low-rank matrices and Next, use the instruction fine-tuning dataset obtained in step (4) to fine-tune and train the large language model through the backpropagation algorithm. During the training process, keep the original weight parameters of the large language model unchanged and only update the low-rank matrix and . This preserves the semantic understanding ability of the large language model and its ability to analyze context and sentiment, enabling the large language model to specifically learn the distribution characteristics and potential structures in the data and improving the model's prediction ability in the case where the car purchase-related information in the input data is relatively sparse.

[0101] After completing the fine-tuning training, save the updated model weights to obtain a user car purchase intention recognition model for subsequent identification of potential car purchase users.

[0102] In step (6), use the large language model after the fine-tuning training in step (5) as the user car purchase intention recognition model, and design a series of prompt instructions to identify potential users with the intention of purchasing a car.

[0103] Specifically, assume that the basic information of user A is as follows: {Age: 37 years old; Gender: female; Education: Bachelor's degree;...}. Then, similar to the construction method of the instruction fine-tuning samples used when fine-tuning and training the large language model, construct prompt instructions such as the following example:

[0104] "Please assess the likelihood of this individual purchasing a car in the near future, presented on a scale of 0-1 and briefly explain the reason of this score: Gender: female; Age: 37; Education: Bachelor's degree...". Input the prompt instructions into the user car purchase intention recognition model, and the user car purchase intention recognition model will output a score in the range of 0-1 and the corresponding basis for the score.

[0105] Similarly, after the user car purchase intention recognition model outputs the scores of all users to be analyzed, the top N users with the highest scores or users with scores greater than the threshold can be selected through sorting as the target customer group for the marketing of auto installment products.

[0106] Then, construct an auto product recommendation model through steps (7) to (9) below.

[0107] In step (7), construct an auto product recommendation model.

[0108] According to some embodiments of the present invention, when constructing a car product recommendation model, in order to achieve accurate recommendation according to three attributes of price range, energy type, and category when recommending cars to users, a car attribute classifier can be constructed respectively according to price, energy type, and category, which is specifically used for the classification of corresponding attributes. Among them, each attribute of the car product is recommended using a dedicated car attribute classifier, which can improve the accuracy of recommendation and overcome the problem that the relatively sparse relevant information of users' car purchases in the banking system affects the accuracy of recommendation.

[0109] Furthermore, the car attribute classifier may further include multiple sub-classifiers, where different sub-classifiers are constructed using different algorithm models.

[0110] For example, for each car attribute classifier, a sub-classifier based on Gradient Boosting Decision Tree (GBDT), a sub-classifier of Adaptive Boosting (AdaBoost), and a sub-classifier based on Multi-Layer Perceptron (MLP) are constructed respectively. Each sub-classifier learns the mapping relationship between user features and car product attributes independently. After all sub-classifiers are trained, a voting mechanism can be used to comprehensively integrate the category information of the attributes output by the multiple sub-classifiers included in each car attribute classifier, so as to determine an attribute of the car that the user may purchase, further overcoming the problem that the relatively sparse relevant information of users' car purchases in the banking system affects the accuracy of recommendation, and further improving the accuracy of recommendation through the integration of the results of multiple sub-classifiers, enhancing the user experience, and reducing the information interference caused by inaccurate recommendations to users. In this way, for the three classification tasks of price range, fuel type, and category, three sub-classifiers can be constructed respectively, and a total of nine sub-classifiers are constructed in this way.

[0111] Next, in this embodiment, the complete auto installment dataset obtained in step (1) can be used to train each sub-classifier. During training, the complete auto installment dataset can be divided by time and quantity. For example, 20% of the data with more recent time can be used as the cross-period validation set, and the remaining data as the training set. The generalization degree of the model for new trends can be evaluated by verifying the performance of the trained model on the cross-period validation set, and then the hyperparameters of the model can be adjusted. For the parameters related to model training, for example, the maximum number of iterations of the GBDT model and the AdaBoost model can be set to 100, the number of layers of the MLP is 2 layers, the leaky rectified linear unit (ReLU) is used as the activation function, the dimension of the hidden layer can be set to 256, and the maximum number of iterations can be set to 20 times. After training is completed on the training set, the selection of the hyperparameters of the model is evaluated by the performance on the cross-period validation set, and the best set of hyperparameters is fixed. The cross-period validation set is also used for model training to obtain the corresponding sub-classifiers; at the same time, the accuracy, recall, precision, and F1-score of each model on the cross-period validation set are recorded.

[0112] Step (8), associate the target customer group obtained in step (5) with the input features of the auto product recommendation model trained in step (7) to obtain the target customer group dataset.

[0113] The target customer group dataset is respectively input into the nine sub-classifiers trained in step (7), and the prediction results of each sub-classifier are collected. For each attribute, a voting mechanism is adopted to synthesize the prediction results of three sub-classifiers to achieve joint prediction. For example, for an attribute, if the prediction results of two or three sub-classifiers are the same, then this prediction result is used as the prediction result of this attribute; if the results output by the three sub-classifiers are all different, the prediction result of a sub-classifier model with the highest F1-score in the cross-period validation set can be selected as the prediction result of this attribute. In addition, by continuously monitoring the recommendation effect, the weights of each model in the voting mechanism can be dynamically adjusted to ensure the accuracy and timeliness of the recommendation results.

[0114] Step (9), obtain the prediction results of the auto products predicted for each user in step (8) on three attributes, and match them in the vehicle model database, and select the vehicle products that meet the requirements. For example, the information of the top 5 vehicle models with the highest market score and / or the top 5 vehicle models with the highest market sales volume is used as the result of personalized recommendation and pushed to the corresponding users.

[0115] In practical applications, the effects of each model are evaluated according to business feedback, and the model parameters and recommendation strategies are continuously optimized.

[0116] Figure 6The flowchart of the recommendation method for automotive products according to another embodiment of the present invention is schematically shown, where Figure 6 It is the overall flowchart of the model application stage, describing the overall calculation framework of the application stage.

[0117] Refer to Figure 6 , in the application stage, the purchase intention score of the target user whose purchase intention needs to be evaluated can be identified first by using the user purchase intention recognition model, and then the users with high purchase intention can be screened out according to the potential purchase conditions. Then, the specific purchase tendency of the users with high purchase intention is predicted by using the automotive product recommendation model and recommendations are made.

[0118] Among them, the initial input of the application stage is the initial data of the target user in the banking system. According to the highly important features screened out by the feature engineering and feature analysis module in the training stage, the corresponding features of the target user are extracted from the initial data, and then a prompt instruction is generated. Then, the prompt instruction is input into the user purchase intention recognition model, and the purchase intention recognition result (such as a score) is given by the user purchase intention recognition model. Then, when the purchase intention recognition result of the target user meets the potential purchase conditions, the target user is determined as a user with high purchase intention. Next, the feature data of this user can be input into the automotive product recommendation model after being sorted according to the input data features of the automotive product recommendation model to obtain the specific purchase tendency.

[0119] Figure 7 The block diagram of the recommendation device for automotive products according to an embodiment of the present invention is schematically shown.

[0120] As Figure 7 shown, the recommendation device 700 for automotive products in this embodiment may include a first acquisition module 710, a feature engineering and feature analysis module 720, a purchase intention prediction module 730, and an automotive product recommendation module 740.

[0121] The first acquisition module 710 is used to obtain the authorization of the first user to use the data in the banking system, and obtain the data of the first user in the banking system after obtaining the authorization of the first user, so as to obtain the first initial data. In one embodiment, the first acquisition module 710 may perform operation S301 and operation S302 introduced above.

[0122] The feature engineering and feature analysis module 720 is used to screen out M feature data corresponding to M features from the first initial data to obtain a first feature set, where the M features are a set of features most relevant to the user's purchase intention screened out by the feature engineering algorithm, and M is an integer greater than or equal to 2. In one embodiment, the feature engineering and feature analysis module 720 may perform operation S303 introduced above.

[0123] The car purchase intention prediction module 730 is used to construct a first prompt instruction represented in natural language by using the first feature set, input the first prompt instruction into the user car purchase intention recognition model, and obtain the first car purchase intention recognition result output by the user car purchase intention recognition model, where the user car purchase intention recognition model is obtained by fine-tuning a pre-trained large language model. In one embodiment, the car purchase intention prediction module 730 may perform the operations S304 and S305 introduced above.

[0124] The automotive product recommendation module 740 is used to recommend automotive products to the first user when the first car purchase intention recognition result meets the potential car purchase conditions. In one embodiment, the automotive product recommendation module 740 may perform the operation S306 introduced above.

[0125] In one embodiment, the feature engineering and feature analysis module 720 is specifically used to train a random forest model to predict the user's car purchase intention based on the user's data in the banking system after obtaining the user's authorization to use their data in the banking system; and after the random forest model is trained, based on the importance of the input features of the random forest model and the Shapley Additive Explanation (SHAP) values, M features are selected. Among them, the feature engineering algorithm is constructed using a random forest model.

[0126] In one embodiment, the feature engineering and feature analysis module 720 is further used to: obtain the authorization of at least one car purchaser to use their data in the banking system, where the car purchaser is a user who has purchased an auto installment product in the banking system; after obtaining the authorization of the car purchaser, obtain the car purchaser's data in the banking system, and obtain n labels of the car purchaser based on the category information of n attributes of the car in the auto installment product purchased by the car purchaser, where n is an integer greater than or equal to 2; train n random forest models, each random forest model is used to predict one of the n attributes; where, when training each random forest model, the car purchaser is labeled with the label corresponding to the attribute to be predicted by the random forest model, and the car purchaser's data in the banking system is used as the input to train the random forest model; after the n random forest models are trained, based on the importance of the input features of the n random forest models and the SHAP values, M features are selected.

[0127] In one embodiment, the automotive product recommendation module 740 is further configured to input the first initial data into an automotive product recommendation model, and obtain the recommended category information of automotive products output by the automotive product recommendation model; wherein, the automotive product recommendation model is trained by using the data of the purchased vehicle users in the banking system after obtaining the authorization of at least one purchased vehicle user to use the data in their banking system, and the purchased vehicle users are users who have purchased automotive installment products in the banking system; and recommend the automotive products indicated by the recommended category information to the first user.

[0128] In one embodiment, the automotive product recommendation module 740 is further configured to: when the automotive product recommendation model includes n automotive attribute classifiers corresponding one-to-one to n attributes, and each automotive attribute classifier includes multiple sub-classifiers, and different sub-classifiers are constructed by using different algorithm models, adopt a voting mechanism to comprehensively obtain the category information of the attributes output by the multiple sub-classifiers included in each automotive attribute classifier, so as to obtain the category information of the attributes corresponding to each automotive attribute classifier.

[0129] The recommendation device 700 can execute the Figures 3 to 6 automotive product recommendation method described in the reference, for details, please refer to the foregoing introduction, which will not be elaborated herein.

[0130] Figure 8 The block diagram of the training device of the user's car purchase intention recognition model according to an embodiment of the present invention is schematically shown.

[0131] As Figure 8 shown, the training device 800 according to this embodiment may include a second acquisition module 810, a feature engineering and feature analysis module 720, and a fine-tuning training module 830.

[0132] The second acquisition module 810 is configured to obtain the authorization of the second user to use the data in their banking system, and after obtaining the authorization of the second user, obtain the data of the second user in the banking system to obtain the second initial data, where the second user is a purchased vehicle user who has purchased an automotive installment product in the banking system or an unpurchased vehicle user who has not used any automotive-related products in the banking system. In one embodiment, the second acquisition module 810 may execute the operations S401 and S402 described above.

[0133] In the training device 800, the feature engineering and feature analysis module 720 is specifically configured to: screen out M feature data corresponding to M features from the second initial data to obtain a second feature set; when the second user is a user who has purchased a vehicle, use the information of having purchased a vehicle as a label to annotate the second feature set to obtain a positive sample data; and when the second user is a user who has not purchased a vehicle, use the information of not having purchased a vehicle as a label to annotate the second feature set to obtain a negative sample data. Wherein, the M features are a group of features screened out by a feature engineering algorithm and most relevant to the user's vehicle purchase intention, and M is an integer greater than or equal to 2. In one embodiment, the feature engineering and feature analysis module 720 may perform operations S403 and S404 introduced above.

[0134] The fine-tuning training module 830 is configured to, after obtaining an equal amount of positive sample data and negative sample data, construct a prompt instruction represented in natural language for each piece of positive sample data and each piece of negative sample data respectively to obtain an instruction fine-tuning data set, and use the instruction fine-tuning data set to fine-tune and train a large language model to obtain a user vehicle purchase intention recognition model. In one embodiment, the fine-tuning training module 830 may perform operations S405 and S406 introduced above.

[0135] In one embodiment, the training device 800 may be integrated into the recommendation device 700. In other embodiments, the training device 800 and the recommendation device 700 may be deployed separately, wherein both the recommendation device 700 and the training device 800 may call the feature engineering and feature analysis module 720 to perform corresponding operations.

[0136] According to an embodiment of the present invention, the training device 800 may execute the training method of the user vehicle purchase intention recognition model described in Figures 4 to 6 For specific reference to the above introduction, it will not be elaborated here.

[0137] According to an embodiment of the present invention, any plurality of modules among the first acquisition module 710, the feature engineering and feature analysis module 720, the car purchase intention prediction module 730, the automotive product recommendation module 740, the second acquisition module 810, and the fine-tuning training module 830 may be combined and implemented in one module, or any one of them may be split into multiple modules. Or, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the first acquisition module 710, the feature engineering and feature analysis module 720, the car purchase intention prediction module 730, the automotive product recommendation module 740, the second acquisition module 810, and the fine-tuning training module 830 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or may be implemented by any other reasonable means such as hardware or firmware by integrating or packaging circuits, or may be implemented in any one of the three implementation manners of software, hardware, and firmware, or in any suitable combination of several of them. Or, at least one of the first acquisition module 710, the feature engineering and feature analysis module 720, the car purchase intention prediction module 730, the automotive product recommendation module 740, the second acquisition module 810, and the fine-tuning training module 830 may be at least partially implemented as a computer program module, and when the computer program module is run, it can execute the corresponding functions.

[0138] Figure 9 A block diagram of an electronic device according to an embodiment of the present invention is schematically shown.

[0139] As Figure 9 shown, the electronic device 900 according to an embodiment of the present invention includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage section 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general microprocessor (such as a CPU), an instruction set processor and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), and so on. The processor 901 may also include on-board memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0140] In the RAM 903, various programs and data required for the operation of the electronic device 900 are stored. The processor 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. The processor 901 performs various operations of the method flow according to the embodiments of the present invention by executing the programs in the ROM 902 and / or the RAM 903. It should be noted that the programs may also be stored in one or more memories other than the ROM 902 and the RAM 903. The processor 901 may also perform various operations of the method flow according to the embodiments of the present invention by executing the programs stored in the one or more memories.

[0141] According to an embodiment of the present invention, the electronic device 900 may further include an input / output (I / O) interface 905, and the input / output (I / O) interface 905 is also connected to the bus 904. The electronic device 900 may further include one or more of the following components connected to the input / output (I / O) interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage portion 908 including a hard disk, etc.; and a communication portion 909 including a network interface card such as a LAN card, a modem, etc. The communication portion 909 performs communication processing via a network such as the Internet. A driver 910 is also connected to the input / output (I / O) interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the driver 910 as needed so that a computer program read from it can be installed into the storage portion 908 as needed.

[0142] The present invention also provides a computer-readable storage medium, which may be included in the device / device / system described in the above embodiments; or may exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of the present invention is implemented.

[0143] An embodiment of the present invention further includes a computer program product, which includes a computer program, and the computer program includes program codes for executing the method shown in the flowchart. When the computer program product runs in a computer system, the program codes are used to cause the computer system to implement the method provided by the embodiments of the present invention.

[0144] When the computer program is executed by the processor 901, the above functions defined in the system / device according to the embodiments of the present invention are executed. According to an embodiment of the present invention, the above-described systems, devices, modules, units, etc. may be implemented by computer program modules.

[0145] The embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present invention.

Claims

1. A method for recommending automobile products, characterized in that: The method comprises: Obtaining the first user's authorization to use his / her data in the banking system; After obtaining authorization from the first user, obtaining the data of the first user in the bank system to obtain first initial data; M feature data corresponding to M features are screened out from the first initial data to obtain a first feature set, wherein the M features are a set of features most relevant to the user's car purchase intention screened out by a feature engineering algorithm, wherein M is an integer greater than or equal to 2; wherein the feature engineering algorithm is constructed by n random forest models, each random forest model is used to predict one of n attributes of the car, and n is an integer greater than or equal to 2; constructing a first prompt instruction expressed in natural language using the first feature set; Inputting the first prompt instruction to a user car purchase intention recognition model, and obtaining a first car purchase intention recognition result output by the user car purchase intention recognition model, wherein the user car purchase intention recognition model is obtained by fine-tuning a pre-trained large language model; and When the first car purchase intention identification result meets the potential car purchase conditions, recommending a car product to the first user; The process of fine-tuning the pre-trained large language model to obtain the user car purchase intention recognition model includes: After obtaining the user's authorization to use the data in the bank system, the data of the users who have purchased the car and the data of the users who have not purchased the car are obtained, and M feature data corresponding to the M features are respectively selected therefrom to construct an instruction fine-tuning data set, The large language model is fine-tuned and trained using the instruction fine-tuning data set to obtain the user car purchase intention recognition model, wherein the car-purchasing users are users who have purchased car installment products in the bank system, and the non-car-purchasing users are users who have not used any car-related products in the bank system; The screening process of the M features includes: Obtaining the authorization of the car-purchasing user to use the data in his / her bank system; After obtaining authorization from the user who has purchased the car, the data of the user who has purchased the car in the bank system is obtained, and n tags of the user who has purchased the car are obtained based on the category information of n attributes of the car in the car installment product purchased by the user who has purchased the car; Training n random forest models, wherein when training each random forest model, the user who has purchased the car is labeled with a label corresponding to the attribute to be predicted by the random forest model, and the data of the user who has purchased the car in the bank system is used as input to train the random forest model; and After the training of the n random forest models is completed, the M features are screened out based on the feature importance of the n random forest models.

2. The method according to claim 1, characterized in that The screening process of the M features specifically includes: After the training of the n random forest models is completed, the M features are screened out based on the importance of the input features and the Shapley additivity interpretation SHAP value of the n random forest models.

3. The method according to claim 1, characterized in that When the first car purchase intention identification result meets the potential car purchase condition, recommending a car product to the first user includes: Inputting the first initial data into an automobile product recommendation model, and obtaining recommended category information of automobile products output by the automobile product recommendation model; wherein the automobile product recommendation model is trained using the data of at least one automobile purchasing user in the bank system after obtaining authorization from at least one automobile purchasing user to use the data in the bank system; The automobile product indicated by the recommendation category information is recommended to the first user.

4. The method according to claim 3, characterized in that The recommendation category information includes category information of the n attributes, and the automobile product recommendation model includes n automobile attribute classifiers corresponding to the n attributes one by one.

5. The method according to claim 4, characterized in that Each of the automobile attribute classifiers includes a plurality of sub-classifiers, wherein different sub-classifiers are constructed using different algorithm models, wherein the step of obtaining the recommended category information of automobile products output by the automobile product recommendation model includes: The category information of the attributes output by the multiple sub-classifiers included in each of the automobile attribute classifiers is synthesized by a voting mechanism to obtain the category information of the attributes corresponding to each of the automobile attribute classifiers.

6. The method according to claim 1, characterized in that The n attributes include at least one of the following: price, category, and energy type.

7. A training method for a user car purchase intention recognition model based on a large language model, characterized in that: The method comprises: Obtaining authorization from a second user for using data in the bank system, where the second user is a user who has purchased a car installment product in the bank system or a user who has not purchased a car but has not used any car-related products in the bank system; After obtaining authorization from the second user, acquiring the data of the second user in the bank system to obtain second initial data; M feature data corresponding to the M features are screened out from the second initial data to obtain a second feature set, wherein the M features are a set of features most relevant to the user's car purchasing intention screened out by a feature engineering algorithm, wherein M is an integer greater than or equal to 2; wherein the feature engineering algorithm is constructed by n random forest models, each random forest model is used to predict one of the n attributes of the car, and n is an integer greater than or equal to 2; When the second user is the user who has purchased a car, the second feature set is labeled with the information of the purchased car as a label to obtain a positive sample data; and when the second user is the user who has not purchased a car, the second feature set is labeled with the information of the non-purchased car as a label to obtain a negative sample data; After obtaining an equal amount of the positive sample data and the negative sample data, constructing a prompt instruction expressed in natural language for each piece of the positive sample data and each piece of the negative sample data, respectively, to obtain an instruction fine-tuning data set; and Using the instruction fine-tuning data set to fine-tune the large language model, so as to obtain the user car purchase intention recognition model; The screening process of the M features includes: After obtaining authorization from the user who has purchased the car, the data of the user who has purchased the car in the bank system is obtained, and n tags of the user who has purchased the car are obtained based on the category information of n attributes of the car in the car installment product purchased by the user who has purchased the car; Training n random forest models, wherein when training each random forest model, the user who has purchased the car is labeled with a label corresponding to the attribute to be predicted by the random forest model, and the data of the user who has purchased the car in the bank system is used as input to train the random forest model; and After the training of the n random forest models is completed, the M features are screened out based on the feature importance of the n random forest models.

8. A device for recommending automobile products, characterized in that: The device comprises: A first acquisition module, used to obtain the authorization of the first user to use the data in the bank system, and to obtain the data of the first user in the bank system after obtaining the authorization of the first user, so as to obtain first initial data; A feature engineering and feature analysis module, configured to filter out M feature data corresponding to M features from the first initial data to obtain a first feature set, wherein the M features are a set of features most relevant to the user's car purchase intention filtered out by a feature engineering algorithm, wherein M is an integer greater than or equal to 2; wherein the feature engineering algorithm is constructed by n random forest models, each random forest model is used to predict one of n attributes of the car, and n is an integer greater than or equal to 2; a car purchase intention prediction module, configured to construct a first prompt instruction expressed in natural language using the first feature set, input the first prompt instruction to a user car purchase intention recognition model, and obtain a first car purchase intention recognition result output by the user car purchase intention recognition model, wherein the user car purchase intention recognition model is obtained by fine-tuning a pre-trained large language model; and An automobile product recommendation module, configured to recommend an automobile product to the first user when the first automobile purchase intention recognition result meets the potential automobile purchase conditions; The process of fine-tuning the pre-trained large language model to obtain the user car purchase intention recognition model includes: After obtaining the user's authorization to use the data in the bank system, the data of the users who have purchased the car and the data of the users who have not purchased the car are obtained, and M feature data corresponding to the M features are respectively selected therefrom to construct an instruction fine-tuning data set, The large language model is fine-tuned and trained using the instruction fine-tuning data set to obtain the user car purchase intention recognition model, wherein the car-purchasing users are users who have purchased car installment products in the bank system, and the non-car-purchasing users are users who have not used any car-related products in the bank system; Wherein, the feature engineering and feature analysis modules are also used for: Obtaining the authorization of the car-purchasing user to use the data in his / her bank system; After obtaining authorization from the user who has purchased the car, the data of the user who has purchased the car in the bank system is obtained, and n tags of the user who has purchased the car are obtained based on the category information of n attributes of the car in the car installment product purchased by the user who has purchased the car, where n is an integer greater than or equal to 2; Training n random forest models, wherein when training each random forest model, the user who has purchased the car is labeled with a label corresponding to the attribute to be predicted by the random forest model, and the data of the user who has purchased the car in the bank system is used as input to train the random forest model; and After the training of the n random forest models is completed, the M features are screened out based on the feature importance of the n random forest models.

9. A training device for a user car purchase intention recognition model based on a large language model, characterized in that: The training device comprises: A second acquisition module is used to obtain the second user's authorization to use the data in the bank system, and after obtaining the second user's authorization, obtain the second user's data in the bank system to obtain second initial data; wherein the second user is a user who has purchased a car installment product in the bank system or a user who has not purchased a car but has not used any car-related products in the bank system; Feature engineering and feature analysis modules for: M feature data corresponding to the M features are screened out from the second initial data to obtain a second feature set; wherein the M features are a set of features most relevant to the user's car purchase intention screened out by a feature engineering algorithm, wherein M is an integer greater than or equal to 2; wherein the feature engineering algorithm is constructed by n random forest models, each random forest model is used to predict one of the n attributes of the car, and n is an integer greater than or equal to 2; When the second user is the user who has purchased a car, the second feature set is labeled with the information of the purchased car as a label to obtain a positive sample data; and when the second user is the user who has not purchased a car, the second feature set is labeled with the information of the non-purchased car as a label to obtain a negative sample data; A fine-tuning training module is used for constructing prompt instructions expressed in natural language for each piece of the positive sample data and each piece of the negative sample data after obtaining an equal amount of the positive sample data and the negative sample data, so as to obtain an instruction fine-tuning data set, and fine-tuning the large language model using the instruction fine-tuning data set to obtain the user car purchase intention recognition model; Wherein, the feature engineering and feature analysis modules are also used for: Obtaining the authorization of the car-purchasing user to use the data in his / her bank system; After obtaining authorization from the user who has purchased the car, the data of the user who has purchased the car in the bank system is obtained, and n tags of the user who has purchased the car are obtained based on the category information of n attributes of the car in the car installment product purchased by the user who has purchased the car, where n is an integer greater than or equal to 2; Training n random forest models, wherein when training each random forest model, the user who has purchased the car is labeled with a label corresponding to the attribute to be predicted by the random forest model, and the data of the user who has purchased the car in the bank system is used as input to train the random forest model; and After the training of the n random forest models is completed, the M features are screened out based on the feature importance of the n random forest models.

10. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

11. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

12. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images