Product display method, device and related product

Abstract

The application relates to the fields of machine learning and financial technology, and discloses a product display method and device and related products. The method comprises the following steps: in response to a display instruction sent by a client, obtaining product information of a plurality of products to be displayed, user portrait data, user behavior data and product access scene data of a current user; wherein the product access scene data comprises at least one of the following: a client type and an access time; preprocessing the product information of the plurality of products to be displayed, the user portrait data, the user behavior data and the product access scene data of the current user to obtain a processing result; generating a product ranking result by using a pre-trained product ranking model according to the processing result; and displaying the plurality of products to be displayed on a front-end page of the client based on the product ranking result. The application can integrate product information and user multi-dimensional demand data, generate a product ranking result that meets the personalized demand of a user, and effectively improve the click rate and conversion rate of products.

Description

Technical Field

[0001] This application relates to the fields of machine learning and fintech, and more specifically, to a product display method, apparatus, and related products. Background Technology

[0002] Currently, on various internet platforms, users can quickly learn about different products through product lists displayed on the front-end page, and then make decisions to browse, select, or purchase. Therefore, the rationality of the product display method is directly related to user experience and the platform's conversion efficiency.

[0003] Generally, most platforms on the market rely on certain static indicators, such as product launch time and price, to sort and display products. Taking the existing non-motor insurance mall system as an example, its insurance product list is displayed in this way, that is, by sorting and displaying products according to indicators such as product launch time, price, or coverage amount.

[0004] This product display method fails to meet the personalized needs of different users, resulting in low accuracy in the displayed product list, which negatively impacts click-through rates and conversion rates. Therefore, it is urgent to solve this technical problem. Summary of the Invention

[0005] In view of the above situation, this application provides a product display method, device and related products, which aim to solve the technical problem of low accuracy of product list adaptation in the prior art.

[0006] In a first aspect, embodiments of this application provide a product display method, the method comprising: In response to a display command sent by the client, the system acquires product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; The product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data are preprocessed to obtain the processing result; Using a pre-trained product ranking model, product ranking results are generated based on the processing results; Based on the product sorting results, the multiple products to be displayed are shown on the front-end page of the client.

[0007] Secondly, embodiments of this application also provide a product display device, the device comprising: The acquisition module is used to respond to the display command sent by the client and acquire product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; The preprocessing module is used to preprocess the product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data to obtain the processing results. The sorting module is used to generate product sorting results based on the processing results using a pre-trained product sorting model. The display module is used to display the multiple products to be displayed on the front-end page of the client based on the product sorting results.

[0008] Thirdly, embodiments of this application also provide an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the product demonstration method described above.

[0009] Fourthly, embodiments of this application also provide a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the steps of the product display method described above.

[0010] By utilizing the above technical solutions, the product display method, apparatus, and related products provided in this application embodiment trigger a real-time data acquisition process in response to the client's display command. The acquired latest product information provides key feature data to characterize the product. The current user profile data can characterize the user's core attributes to locate basic needs (such as information like "28-year-old woman," which indicates that she may be interested in women-specific insurance). The latest user behavior data can uncover the user's dynamic and real preferences (such as browsing medical insurance or clicking on combination insurance, which reflects the user's preference for health insurance). Product access scenario data can accurately capture the user's immediate needs (such as accessing through an APP or browsing on weekends, which means that the user may have ample time to compare insurance options). These three types of user data and product information form a multi-dimensional complementary data source. Preprocessing of these data is then performed. It can effectively purify data noise, unify data format, and improve data usability, providing high-quality input for subsequent model calculations. The pre-trained product ranking model can deeply integrate product information and multi-dimensional user demand data, breaking through the limitations of existing technologies that rely solely on static product indicators. Through algorithms, it generates list-level product ranking results that fit individual user differences, behavioral habits, and current scenarios. Finally, based on this ranking result, products are displayed on the client-side front-end page, effectively solving the problems of existing display methods failing to meet the personalized needs of different users and having low accuracy. This allows products with high user intent to be presented first, aligning with users' actual preferences, attributes, and current scenarios, guiding users to quickly find products that meet their needs, effectively improving product click-through rates and conversion rates, and enhancing user experience and the platform's core competitiveness.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This illustration shows an application environment diagram of the product display method provided in an embodiment of this application; Figure 2 A flowchart illustrating the product display method provided in an embodiment of this application is shown; Figure 3 A schematic diagram of the structure of the product display device provided in an embodiment of this application is shown; Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0016] As mentioned earlier, most platforms on the market typically rely on static metrics, such as product launch time and price, to sort and display products. For example, existing non-motor vehicle insurance e-commerce systems use this method to display their insurance product lists, sorting and displaying products by launch time, price, or coverage amount. However, this product display method fails to meet the personalized needs of different users, resulting in low accuracy in the displayed product list and consequently affecting click-through rates and conversion rates. Therefore, this invention proposes a product display method, apparatus, and related products, which will be described in detail below through specific embodiments.

[0017] To facilitate understanding of this embodiment, the product display method disclosed in this application embodiment will first be described in detail. The product display method provided in this application embodiment can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The server can respond to display commands sent by the client by obtaining product information for multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data. The product access scenario data includes at least one of the following: client type and access time. The server preprocesses the product information, user profile data, user behavior data, and product access scenario data to obtain a processing result. Using a pre-trained product ranking model, the server generates a product ranking result based on the processing result. Based on the product ranking result, the server displays the multiple products to be displayed on the client's front-end page. This invention triggers a real-time data acquisition process in response to client display commands. The acquired latest product information provides key feature data to characterize the product, current user profile data can characterize the user's core attributes to locate basic needs, the latest user behavior data can uncover the user's dynamic and real preferences, and product access scenario data can accurately capture the user's immediate needs. These three types of user data and product information form a multi-dimensional complementary data source. Preprocessing this data can effectively purify data noise, unify data format, and improve data usability, providing high-quality input for subsequent model calculations. The pre-trained product ranking model can deeply integrate product information and multi-dimensional user demand data, breaking through the limitations of existing technologies that rely solely on static product indicators. Through algorithms, it generates list-level product ranking results that fit individual user differences, behavioral habits, and current scenarios. Finally, based on this ranking result, products are displayed on the client's front-end page, effectively solving the problems of existing display methods failing to meet the personalized needs of different users and having low accuracy. This allows products with high user intent to be presented first, aligning with users' actual preferences, attributes, and current scenarios, guiding users to quickly find products that meet their needs, effectively improving product click-through rates and conversion rates, and enhancing user experience and the platform's core competitiveness. The client device can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. In some possible implementations, this method can be implemented by the processor calling computer-readable instructions stored in memory. The invention will now be described in detail through specific embodiments.

[0018] Figure 2 This document illustrates a flowchart of a product display method provided in an embodiment of this application. Figure 2 It can be seen that the embodiments of this application include at least steps S201-S204: S201: In response to the display command sent by the client, obtain product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; S202: Preprocess the product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data to obtain the processing result; S203: Using a pre-trained product ranking model, generate product ranking results based on the processing results; S204: Based on the product sorting results, display the multiple products to be displayed on the front-end page of the client.

[0019] As can be seen, this application embodiment triggers a real-time data acquisition process by responding to the client's display command. The latest product information acquired provides key feature data to characterize the product. The current user profile data can characterize the user's core attributes to locate basic needs (such as information like "28-year-old woman," which indicates that she may be interested in women-specific insurance). The latest user behavior data can uncover the user's dynamic and real preferences (such as browsing medical insurance or clicking on combination insurance, which reflects the user's preference for health insurance). Product access scenario data can accurately capture the user's immediate needs (such as accessing through an APP or browsing on weekends, which means that the user may have ample time to compare insurance options). These three types of user data and product information form a multi-dimensional complementary data source. Preprocessing these data can effectively purify data noise and unify data... By formatting and improving data usability, high-quality input is provided for subsequent model calculations. The pre-trained product ranking model can deeply integrate product information and multi-dimensional user demand data, breaking through the limitations of existing technologies that rely solely on static product indicators. Through algorithms, it generates list-level product ranking results that fit individual user differences, behavioral habits, and current scenarios. Finally, based on this ranking result, products are displayed on the client-side front-end page, effectively solving the problems of existing display methods failing to meet the personalized needs of different users and having low accuracy. This allows products with high user intent to be presented first, aligning with users' actual preferences, attributes, and current scenarios, guiding users to quickly find products that meet their needs, effectively improving product click-through rates and conversion rates, and enhancing user experience and the platform's core competitiveness.

[0020] The steps S201-S204 described above will be explained in detail below.

[0021] Regarding the above S201-S202: It should be noted that the product display method provided in this application embodiment can be applied to product sorting and display tasks in any vertical field within an industry. In this embodiment and the steps of the following embodiments, the technical solution in this application is described using the display task of insurance products as an example, but this does not constitute a specific limitation on the application scenarios to which the technical solution in this application is applicable.

[0022] First, the execution entity in this embodiment receives the display instruction sent by the client. The client here can be, for example, an APP, H5, or a WeChat mini-program. Here, APP stands for "Application," specifically referring to a native application directly installed on the operating system of mobile devices such as phones and tablets, possessing an independent running environment and system permissions; H5 stands for "HTML5," essentially an application developed based on web page technology, requiring no separate download and installation, and accessible via a mobile browser (such as WeChat Browser, Safari, or Chrome).

[0023] The display command is triggered by the user's active operation on the client. For example, when a user opens a non-motor insurance product page in the insurance mall in the APP, refreshes the product list page on the H5 page, or jumps to the product list page from other pages in various channels, the client will detect these operations and generate corresponding display commands, and then send the commands to the execution subject of this application embodiment.

[0024] In this embodiment of the application, the executing entity responds to the display command and obtains product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data.

[0025] Product information is a structured data set that characterizes the core attributes and key value of a product, including but not limited to: product type, product price, and popularity. For example, if the product is insurance, the product information includes product type (e.g., travel insurance, accident insurance), coverage (e.g., short-term domestic accidental medical expenses), sum insured (e.g., 500,000 RMB accidental death benefit), price (e.g., 120 RMB / policy), and popularity (e.g., high popularity). In practice, this product information can be obtained from the insurance product management system. Popularity can be quantified by real-time statistics of historical and current user access, clicks, and purchase data.

[0026] User profile data is a structured collection of data that characterizes a user's core attributes, including but not limited to: age (e.g., 28 years old), gender (e.g., female), occupation, and place of residence. In practice, it can be obtained by reading user registration information.

[0027] User behavior data refers to data generated during user interactions with a product, including but not limited to: browsing activity (e.g., viewing details of three travel insurance products), click activity (e.g., clicking the "Insurance Product 1" button), purchase activity (e.g., successfully purchasing home insurance), dwell time (e.g., spending 2 minutes and 30 seconds on a pet insurance details page), and scrolling depth (e.g., reaching the bottom of a health insurance details page). This data directly reflects users' preferences for different products. In implementation, it can be collected and stored in real time by a user behavior collection system during user interactions with the product.

[0028] Product access scenario data refers to the context-related characteristics of users accessing a product. It is used to capture the real-time context conditions during user access, helping to improve the accuracy of ranking. Product access scenario data includes, but is not limited to: access channels (such as apps, WeChat mini-programs), access time (such as 8 PM on a weekday), etc. In implementation, the client can collect device environment information and access timestamps and upload them to the execution entity of this application embodiment.

[0029] After obtaining the raw data such as product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data, the product information of multiple products, user profile data of the current user, user behavior data, and product access scenario data are preprocessed so that they can be input into the pre-trained product ranking model.

[0030] In practice, the preprocessing steps include: first, data cleaning to remove data with missing key information or outliers (such as records with negative dwell time), and supplementing necessary missing data (such as marking unrecorded user gender as "unknown"); then, standardizing or normalizing numerical features to convert indicators of different magnitudes, such as price, coverage amount, and dwell time, to the 0-1 range to avoid affecting model weights; next, encoding categorical features to convert non-numerical data such as product type, access channel, and gender into computer-recognizable numerical codes (e.g., 1 for APP, 2 for H5); for text-related data (such as product type descriptions), simple word segmentation or feature extraction can be performed; finally, integrating the processed product information, user behavior data, user profile data, and product access scenario data into a structured feature vector in a unified format to ensure compliance with model input requirements.

[0031] Regarding the above S203-204: After obtaining the preprocessed product information of multiple products, the user profile data of the current user, user behavior data, and product access scenario data, the processing results can be input into the pre-trained product ranking model to generate product ranking results; and based on the product ranking results, each product can be displayed on the front-end page of the client.

[0032] In implementation, before generating product ranking results based on the processing results using the pre-trained product ranking model, the product display method further includes: Acquire user profile data, user behavior data, and product access scenario data from multiple sample users; Based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, a dataset is generated; wherein, the operation data includes: no click, click, and purchase; the dataset is used to train an initial product ranking model to generate the pre-trained product ranking model.

[0033] In this embodiment, to generate a pre-trained product ranking model, a dataset needs to be constructed first. Specifically, the steps for generating the dataset are as follows: First, the composition of the raw data for each sample is determined. The raw data for each sample includes user profile data, user behavior data, product access scenario data, product information for each product to be displayed, and user behavior data showing the user's actions towards each product. The product information for each product, the user profile data, user behavior data, and product access scenario data together constitute the feature part of the sample, used to characterize the core attributes and interaction context of the user and the product. The product relevance tags (e.g., purchase = 3, click = 2, no click = 0) for multiple products to be displayed, transformed from the action data, serve as the sample's tags, used to indicate the degree to which the user's needs match the various products to be displayed.

[0034] For example, the user profile data of the sample user includes: a 30-year-old female living in a first-tier city; user behavior data includes: having clicked on the details pages of 2 pet medical insurance products in the past 3 months, with an average dwell time of 80 seconds, and having favorited 1 pet health insurance product; the product access scenario data is that the user accessed the insurance mall via WeChat mini-program at 3 PM on November 2, 2025; the products to be displayed include: A. Pet Medical Insurance (product information: type is pet medical, coverage includes inpatient medical care and vaccination costs, coverage amount of 50,000 yuan, price of 150 yuan / year, high popularity) and B. Home Property Insurance (product information: type is home property insurance, coverage includes the main structure of the house and interior). The sample includes two insurance products: Property Insurance (1 million RMB coverage, 360 RMB / year, moderate popularity) and Short-Distance Travel Insurance (Product Information: Type: Travel Insurance, Coverage includes 3 days of domestic accidental medical treatment, 300,000 RMB coverage, 20 RMB / policy, high popularity). The user data for the sample shows clicks on Pet Medical Insurance (Product A) and no clicks on either Property Insurance (Product B) or Short-Distance Travel Insurance (Product C). Therefore, the corresponding product relevance tags are 2, 0, and 0, respectively. The product information, user profile data, user behavior data, and product access scenario data together constitute the feature part of this sample, while the tags corresponding to the three products together constitute the tag part of this sample.

[0035] Then, all the original sample data are preprocessed. In practice, this may include cleaning samples with missing key features or outliers, encoding classification features (such as product type and access channel), and standardizing or normalizing numerical features (such as price and coverage amount). Finally, the data is integrated to form a standardized dataset with a unified structure that can be directly used for model training.

[0036] After generating the dataset, the model training phase can begin. First, the standardized dataset is divided into training, validation, and test sets according to a preset ratio (e.g., 7:2:1). The training set is used for learning model parameters, the validation set is used to monitor training effectiveness and adjust hyperparameters, and the test set is used to finally evaluate the model's generalization ability. Next, the product ranking model (e.g., LambdaMART, ListNet, ListMLE, etc., list-wise models) is initialized, and the model's hyperparameters (e.g., learning rate, tree depth, number of iterations, etc.) are set. Then, the feature data from the training set is input into the initial model, using sample labels as supervision signals, and the ranking loss function (e.g., based on NDCG normalized cumulative loss) is minimized through backpropagation. The model uses a loss function that combines gain and mean average precision (MAP) to learn the mapping relationship between user features, product features, and tags. During training, the model's ranking performance is evaluated on a validation set after each iteration. If overfitting occurs (e.g., the performance on the validation set decreases while the performance on the training set continues to improve), adjustments are made using regularization, early stopping, and other strategies, while hyperparameters are optimized to improve model performance. When the model's performance on the validation set reaches a preset metric (e.g., the NDCG value stabilizes above 0.8) and no longer improves significantly, training is stopped. Finally, the model's performance is validated on a test set. Once the model is confirmed to have good generalization ability, the pre-trained product ranking model is obtained.

[0037] This embodiment comprehensively collects user profile data, behavioral data, and product access scenario data, combines them with product information to construct a dataset, and trains the initial product ranking model based on this dataset to obtain a pre-trained model. This allows the model to fully learn the real preferences and interaction patterns of different users for various products in different scenarios, accurately capture the adaptation relationship between user needs and product characteristics, and effectively improve the relevance and accuracy of product ranking results. This enables the model to adapt to diverse user groups and access scenarios in subsequent actual product displays, reasonably match potential user needs, reduce invalid product displays, improve user browsing experience and decision-making efficiency, and help increase product click-through conversion rate and purchase conversion rate, thereby optimizing the overall effect of product display and business conversion benefits.

[0038] In some embodiments, the method further includes: Obtain the actual interaction data of the product sorting results, wherein the actual interaction data includes at least one of the following: click data and purchase data; Based on the actual interaction data, determine whether the product ranking result is a high-quality ranking result, and obtain the judgment result; If the judgment result is that the product ranking result is a high-quality ranking result, then new training samples are generated based on the product information, user profile data, user behavior data, product access scenario data, and user operation data of each product in the high-quality ranking result corresponding to the high-quality ranking result. The pre-trained product ranking model is updated using the newly added training samples.

[0039] In this embodiment, after generating multiple product ranking results for the products to be displayed using a pre-trained product ranking model, the actual interaction data between the user and the product ranking results can be obtained. This actual interaction data may include, for example, the number of product clicks and whether a purchase was made.

[0040] After obtaining the actual interaction data for the product ranking results, the system then determines whether the product ranking results are high-quality based on this data. Specifically, the determination method could be as follows: set click-through rate (CTR) and purchase rate (PTR) thresholds; calculate the corresponding CTR, PTR, and other metrics based on the actual interaction data; if the CTR is greater than the CTR threshold, or the PTR threshold is greater than the purchase threshold, then the product ranking result is determined to be high-quality; otherwise, the product ranking result is determined to be low-quality.

[0041] If the product ranking result is a high-quality ranking result, new training samples are generated based on the product information, user profile data, user behavior data, product access scenario data, and user operation data for each product corresponding to the high-quality ranking result. Finally, the pre-trained product ranking model is updated using the new training samples to obtain the updated product ranking model. The specific sample generation steps and model training steps can be found in the aforementioned dataset generation process and model training process, and will not be repeated here.

[0042] This embodiment proposes a product ranking model update scheme. It obtains real user interaction feedback by collecting click and purchase data corresponding to the product ranking results. Then, based on the actual interaction data, it determines high-quality ranking results and filters out effective product ranking results that meet user needs. By integrating product information, user profile data, user behavior data, product access scenario data, and user operation data corresponding to high-quality ranking results, it generates new training samples, enriching the diversity and effectiveness of the model training data. Iterative optimization of the pre-trained product ranking model using these new training samples strengthens the model's ability to capture the relationship between user preferences and product suitability. Ultimately, this embodiment achieves a closed loop where the model continuously learns from actual application results, significantly improving the adaptation accuracy and generalization ability of the product ranking model, making the product ranking results more aligned with users' personalized needs, thereby effectively increasing product click-through rates and conversion rates, and enhancing the platform's core competitiveness.

[0043] In some embodiments, the method further includes: Acquire multimodal data of the multiple products to be displayed; the multimodal data includes product text data, product image data, and product video data; The dataset is generated based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, including: A dataset is generated based on the product information and multimodal data of the multiple products to be displayed, the user profile data of each sample user, the user behavior data, the product access scenario data, and the user's operation data on each of the products to be displayed in the user behavior data; Before the step of generating product ranking results based on the processing results using the pre-trained product ranking model, the method further includes: In response to a display command sent by the client, multimodal data of the multiple products to be displayed are obtained, and each multimodal data is preprocessed to obtain preprocessed multimodal data; the preprocessed multimodal data is then added to the processing result.

[0044] In this embodiment, multimodal data refers to a comprehensive data set presented through various types of information carriers such as text, static images, and dynamic videos, which can comprehensively depict product characteristics from multiple dimensions.

[0045] Taking insurance products as an example, their multimodal data specifically includes insurance product copy, cover images, and video introductions. The insurance product copy, in structured or unstructured text form, clearly presents the product's core attributes (such as type, coverage, sum insured, and price), core selling points (such as claims advantages and underwriting flexibility), and key service information (such as underwriting process and claims processing time). The insurance product cover image, in an intuitive image format (including real-life scene shots and icon designs), displays the product's core identifying elements (such as the product's corresponding scenario, core coverage identifiers, and price / popularity tags). The insurance product video introduction, in the form of short videos, uses a combination of visuals and audio to present key information such as the product's coverage, underwriting process, claims scenarios, and core advantages.

[0046] In this embodiment, the specific process of generating the dataset is as follows: First, the composition of the original data of a single sample is determined. The original data of each sample includes user profile data, user behavior data, product access scenario data of a single sample user, product information and multimodal data of each product to be displayed, and user operation data of the sample user on each product to be displayed in the user behavior data. The product information and multimodal data of each product to be displayed, together with the user profile data, user behavior data and product access scenario data of the sample user, constitute the feature part of the sample. The product relevance labels of multiple products after the operation data is transformed are used as the labels of the sample.

[0047] For example, a sample user, a 30-year-old woman living in a first-tier city, had clicked on the details pages of two pet medical insurance policies within the past three months, spending an average of 80 seconds on each page and having saved one pet health insurance policy. She then accessed the insurance mall via WeChat mini-program at 3 PM on November 2nd, 2025. The products to be displayed included Pet Medical Insurance (Policy A), Home Insurance (Policy B), and Short-Distance Travel Insurance (Policy C). Each product included corresponding product information and multimodal data such as text, images, and video introductions. (The text for Pet Medical Insurance (Policy A) presented its pet medical treatment types, inpatient medical care, and vaccine cost coverage.) The core attributes of the sample include a coverage amount of 50,000 yuan and a price of 150 yuan per year. The cover image showcases pet medical scenarios and trending tags, while the video presentation details the inpatient medical claims process. The user's activity data for Pet Medical Insurance (Type A) shows clicks, while their activity data for Home Insurance (Type B) and Short-Distance Travel Insurance (Type C) shows no clicks. The corresponding product relevance tags are 2, 0, and 0, respectively. These elements together constitute a single sample. Subsequently, all raw sample data underwent preprocessing, such as cleaning samples with missing key features and outliers, and encoding classification features such as product type and access channel. Numerical features such as price and coverage are standardized or normalized. For product text data, noise reduction (removing invalid characters and duplicate information), word segmentation, and stop word filtering are performed first. Then, semantic features are extracted using a pre-trained language model (such as BERT) to generate a fixed-dimensional text feature vector. For product image data, size normalization and pixel value standardization (such as mapping to the 0-1 range) are performed. Data augmentation is performed if necessary through flipping, cropping, etc. Then, visual key features are extracted using CNN or a pre-trained visual model (such as ResNet) and converted into image feature vectors. For product video data, key frames are extracted to reduce redundancy. The above image preprocessing operations are performed on each frame to extract frame-level visual features. Then, inter-frame temporal information is fused using temporal models such as LSTM and Transformer to generate video feature vectors. Finally, the feature vectors of text, image, and video are dimensionally aligned. Weights are assigned according to the importance of each modality to the ranking task. They are then integrated into a unified-dimensional multimodal fusion feature through feature concatenation or cross-modal attention fusion mechanisms. Finally, a standardized dataset with a unified structure that can be directly used for model training is formed.

[0048] During the model application phase, after receiving multimodal data of multiple products to be displayed in response to the display command sent by the client, targeted preprocessing will be carried out for different types of multimodal data. Here, the preprocessing method can refer to the multimodal data processing method in the model training phase. After the preprocessing is completed, the quantized feature vectors corresponding to these multimodal data are fused with other features such as user profile data, user behavior data, and product access scenario data to form a feature input vector of a unified dimension, which is then input into the pre-trained product ranking model.

[0049] This embodiment acquires multimodal data containing text, images, and videos of products to be sorted, and deeply integrates this data with basic product information, sample user profile data, user behavior data, product access scenario data, and operation data of each product to be displayed to generate a dataset. Simultaneously, before inputting the data into the product sorting model, it responds to client display commands, acquires multimodal data of the products to be displayed, preprocesses it, and supplements it into the processing results. This significantly enriches the data dimensions during model training and inference, enabling the dataset to more comprehensively cover product features, user preferences, and scenario attributes. This effectively improves the adaptability of the product sorting model to user needs and access scenarios, thereby generating more accurate product sorting results that fit actual usage scenarios, optimizing the targeting and effectiveness of product display, and ultimately enhancing the user experience.

[0050] During implementation, if the current user's user behavior data and product access scenario data are the same as before the last display, the previous sorting results can be used directly to save computing resources and improve display efficiency.

[0051] In some embodiments, after the step of generating product ranking results based on the processing results using a pre-trained product ranking model, the method further includes: Obtain at least one featured recommended product; Based on the at least one key recommended product and the product ranking result, a new product ranking result is generated.

[0052] In this embodiment, the determination of key recommended products is mainly based on business operation needs, sales targets, or specific promotion strategies. For example, the platform needs to vigorously promote newly launched products during a specific period, blockbuster products to boost quarterly sales performance, products that are prioritized for display as agreed with partners, or exclusive recommended products customized for user lifecycle stages (such as the renewal period of old users). It can also be selected by combining market hotspots (such as travel insurance products corresponding to peak travel periods during holidays), inventory priority, and other rules. These types of products are usually the products that the platform needs to focus on exposing in order to improve conversion or achieve specific business indicators.

[0053] When adding key recommended products to the existing product ranking results, the core logic of the original ranking results will be retained and reasonably integrated. In specific implementation, for example, key recommended products can be inserted into a specified early position in the ranking results (such as the first 3 or the 5th position) according to business rules; alternatively, the positions of some low-priority products can be replaced without significantly affecting the rationality of the original ranking.

[0054] In this embodiment, after the pre-trained model generates product ranking results that fit user preferences, a new ranking is generated by introducing key recommended products and combining them with the original ranking results. This not only preserves the user demand adaptability reflected in the model ranking, but also satisfies specific business requirements through the reasonable integration of key recommended products, thus achieving a balance between user preferences and business goals.

[0055] In some embodiments, after the steps of obtaining product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data, the method further includes: If the current user's user behavior data is not empty, then continue to execute the steps of preprocessing the product information of the multiple products to be displayed, the current user's user profile data, user behavior data, and product access scenario data to obtain the processing results, as well as subsequent steps; If the current user's user behavior data is empty, then a first cold start product ranking result is generated based on the popularity data of each of the products to be displayed; or Based on the user profile data of the current user and other users, calculate the similarity between the current user and other users; determine the closest user based on the similarity scores; and generate the first cold start product ranking result based on the user behavior data of the closest user. Based on the first cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

[0056] In this embodiment, product information for multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data are first obtained. Next, it is determined whether the current user's user behavior data exists (whether it is not empty), and processing is branched accordingly. In the first branch, if the current user's user behavior data is not empty, it means that the current user is an old user with historical interaction records. At this time, the aforementioned steps S202-S204 will continue to be executed.

[0057] In the second branch, if the current user's user behavior data is empty, it indicates that the user is a new user with no historical interaction records. In this case, the first cold start product ranking results will be generated through two optional methods. One method is to directly generate the first cold start ranking results based on the popularity data of each product to be displayed (for example, it can be calculated based on other users' historical clicks, purchases, dwell time, etc. for each product), with products with higher popularity ranking higher. The other method is to first calculate the similarity between the current user and each other user based on the current user's user profile data and the user profile data of other users on the platform (for example, using cosine similarity to calculate the similarity between the processed user profile data), then filter out the closest user with the highest similarity to the current user from the other users, and then obtain the user behavior data of the closest user (i.e., the user's browsing, clicks, purchases, and other interaction records). Based on this behavior data, it is determined that the current user may prefer certain products, and then the product ranking results are generated.

[0058] Finally, based on the initial cold start sorting results, multiple products to be displayed are shown on the client's front-end page.

[0059] This embodiment adapts to both new and existing user scenarios by distinguishing whether user behavior data is empty. Existing users achieve personalized and accurate ranking based on multi-dimensional data preprocessing and subsequent processes. New users generate cold start ranking results through product popularity data, or generate cold start results by matching similar existing user preferences with user profile similarity. This effectively solves the cold start ranking problem for new users who lack behavioral data, while ensuring the accuracy of recommendations for existing users. It takes into account different user needs and platform operation logic, improving the overall product display adaptability and conversion efficiency.

[0060] In some embodiments, if the plurality of products to be displayed includes new products and old products, wherein products not existing in the user behavior data are new products and products existing in the current user's user behavior data are old products, then the current user uses a pre-trained product ranking model to generate a product ranking result based on the processing result, including: The product information of the old products, the user profile data of the current user, the user behavior data, and the product access scenario data are processed and input into the pre-trained product ranking model to generate the product ranking result. The method further includes: Based on the product information of the new product and the old product, determine the product that is closest to the old product; The product sorting results are updated based on the sorting position of the closest old product to generate a second cold start product sorting result; Based on the second cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

[0061] In this embodiment, as mentioned above, user behavior data refers to the current user's past interaction records with platform products, including clicks, purchases, browsing, dwell time, etc. Products that exist in this data are old products (such as accident insurance that the user previously browsed and saved), and products that do not exist in this data are new products (such as newly launched home property insurance that has not been interacted with by users).

[0062] This embodiment employs different sorting strategies for different products. For older products, a pre-trained product sorting model is used to generate an initial sorting result. In the step of generating the sorting result for older products, the input data includes product information of older products, user profile data of the current user, user behavior data, and the preprocessed results of product access scenario data. The above preprocessed multi-dimensional data is input into the sorting model, and the model generates a product sorting result that only includes older products.

[0063] For new products, specifically, the steps for generating the initial ranking position of new products can be implemented by calculating the similarity between the product information of new products and old products (such as using cosine similarity) to find the old products that are most similar to the new products.

[0064] Finally, the product ranking results are updated based on the ranking position of the closest existing product to the old product, generating a second cold start product ranking result. For example, if a new product is highly similar to the old product ranked 3rd in the ranking result, then the initial ranking position of the new product is set to 4th.

[0065] This embodiment distinguishes between old and new products and adopts differentiated sorting logic. Old products are sorted in a personalized manner based on a pre-trained model, ensuring recommendation accuracy. New products determine their initial sorting position based on matching product information with old products, effectively solving the sorting problem of new products without user behavior data support during cold starts. Finally, a complete sorting result is generated through integration and updates, achieving orderly integration of old and new products. This takes into account both users' personalized needs and the promotion needs of new products, improving the rationality of product sorting and users' browsing experience, thereby helping to increase the exposure rate of new products and overall conversion efficiency.

[0066] Those skilled in the art will understand that in the above-described method of the specific embodiments, the order in which the steps are written does not imply a strict execution order, but constitutes no limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0067] It should be noted that in practical applications, all the above-described possible implementation methods can be combined in any way to form possible embodiments of this application, and will not be described in detail here. The information (including but not limited to device information, user information, etc.) and data (including but not limited to data used for analysis, storage, and display) involved in this application are all information and data authorized by the user or fully authorized by all parties. The software tools or components appearing in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0068] Based on the same concept, this application also provides a product display device, which corresponds one-to-one with the product display method in the above embodiments. Figure 3 A schematic diagram of the product display device provided in an embodiment of this application is shown. See also: Figure 3 As shown, the product display device 300 provided in this application embodiment includes: The acquisition module 301 is used to respond to the display command sent by the client and acquire product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; The preprocessing module 302 is used to preprocess the product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data to obtain the processing result; The sorting module 303 is used to generate a product sorting result based on the processing result using a pre-trained product sorting model. The display module 304 is used to display the multiple products to be displayed on the front-end page of the client based on the product sorting results.

[0069] In some embodiments, the apparatus further includes a training module, which, before generating product ranking results based on the processing results using a pre-trained product ranking model, is used to: Acquire user profile data, user behavior data, and product access scenario data from multiple sample users; Based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, a dataset is generated; wherein, the operation data includes: no click, click, and purchase; The initial product ranking model is trained using the dataset to generate the pre-trained product ranking model.

[0070] In some embodiments, the training module is further configured to: Obtain the actual interaction data of the product sorting results, wherein the actual interaction data includes at least one of the following: click data and purchase data; Based on the actual interaction data, determine whether the product ranking result is a high-quality ranking result, and obtain the judgment result; If the judgment result is that the product ranking result is a high-quality ranking result, then new training samples are generated based on the product information, user profile data, user behavior data, product access scenario data, and user operation data of each product in the high-quality ranking result corresponding to the high-quality ranking result. The pre-trained product ranking model is updated using the newly added training samples.

[0071] In some embodiments, in the above-described apparatus, the training module is further configured to: acquire multimodal data of the plurality of products to be displayed; the multimodal data includes product text data, product image data, and product video data; When generating a dataset based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, the training module is specifically used for: A dataset is generated based on the product information and multimodal data of the multiple products to be displayed, the user profile data of each sample user, the user behavior data, the product access scenario data, and the user's operation data on each of the products to be displayed in the user behavior data; Before the step of generating a product ranking result based on the processing result using a pre-trained product ranking model, the acquisition module is further configured to: in response to a display instruction sent by the client, acquire multimodal data of the multiple products to be displayed, preprocess each multimodal data to obtain preprocessed multimodal data, and add the preprocessed multimodal data to the processing result.

[0072] In some embodiments, the apparatus further includes an optimization module, which, after the step of generating product ranking results based on the processing results using a pre-trained product ranking model, is used to: Obtain at least one featured recommended product; Based on the at least one key recommended product and the product ranking result, a new product ranking result is generated.

[0073] In some embodiments, the device further includes a first cold start module, which, after the steps of acquiring product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data, is used to: If the current user's user behavior data is not empty, then continue to execute the steps of preprocessing the product information of the multiple products to be displayed, the current user's user profile data, user behavior data, and product access scenario data to obtain the processing results, as well as subsequent steps; If the current user's user behavior data is empty, then a first cold start product ranking result is generated based on the popularity data of each of the products to be displayed; or Based on the user profile data of the current user and other users, calculate the similarity between the current user and other users; determine the closest user based on the similarity scores; and generate the first cold start product ranking result based on the user behavior data of the closest user. Based on the first cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

[0074] In some embodiments, if the plurality of products to be displayed includes new products and old products, wherein products that do not exist in the current user's user behavior data are new products and products that exist in the current user's user behavior data are old products, then the sorting module is used to input the processing results of the product information of the old products, the user profile data of the current user, user behavior data, and product access scenario data into the pre-trained product sorting model to generate the product sorting result; The device further includes a second cold start module for: Based on the product information of the new product and the old product, determine the product that is closest to the old product; The product sorting results are updated based on the sorting position of the closest old product to generate a second cold start product sorting result; Based on the second cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

[0075] This invention provides a product display device that triggers a real-time data acquisition process in response to a client's display command. The acquired latest product information provides key feature data that characterizes the product. Current user profile data can characterize core user attributes to pinpoint basic needs (e.g., information like "28-year-old woman" indicates potential interest in women-specific insurance). Latest user behavior data can uncover dynamic user preferences (e.g., browsing medical insurance or clicking on combination insurance shows a preference for health insurance). Product access scenario data can accurately capture immediate user needs (e.g., accessing via an app or browsing on weekends suggests ample time for comparison and purchase). These three types of user data, along with product information, form a multi-dimensional complementary data source. Preprocessing this data effectively removes data noise and improves overall performance. The data format is improved, enhancing data usability and providing high-quality input for subsequent model calculations. The pre-trained product ranking model can deeply integrate product information and multi-dimensional user demand data, breaking through the limitations of existing technologies that rely solely on static product indicators. Through algorithms, it generates list-level product ranking results that fit individual user differences, behavioral habits, and current scenarios. Finally, based on this ranking result, products are displayed on the client-side front-end page, effectively solving the problems of existing display methods failing to meet the personalized needs of different users and having low accuracy. This allows products with high user intent to be presented first, aligning with users' actual preferences, attributes, and current scenarios, guiding users to quickly find products that meet their needs, effectively improving product click-through rates and conversion rates, and enhancing user experience and the platform's core competitiveness.

[0076] Specific limitations regarding the product display device can be found in the limitations on the product display method described above, and will not be repeated here. Each module in the aforementioned product display device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0077] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Figure 4 As shown, at the hardware level, this electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may include non-volatile memory, such as at least one disk drive. Of course, this electronic device may also include other hardware required for other business operations.

[0078] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0079] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0080] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a product demonstration device at the logical level. The processor executes the program stored in memory and specifically performs the aforementioned methods.

[0081] The processor may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0082] The electronic device can execute the product display methods provided in several embodiments of this application and be implemented as a product display device. Figure 3 The functions of the embodiments shown are not described in detail here.

[0083] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, enable the electronic device to perform the product demonstration methods provided in various embodiments of this application.

[0084] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0089] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0090] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0091] This application also provides a computer program product that carries program code. The instructions included in the program code can be used to execute the steps of the product display method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0092] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0093] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0094] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A product display method, characterized in that, The method includes: In response to a display command sent by the client, the system acquires product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; The product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data are preprocessed to obtain the processing result; Using a pre-trained product ranking model, product ranking results are generated based on the processing results; Based on the product sorting results, the multiple products to be displayed are shown on the front-end page of the client.

2. The method according to claim 1, characterized in that, Before generating product ranking results based on the processing results using the pre-trained product ranking model, the method further includes: Acquire user profile data, user behavior data, and product access scenario data from multiple sample users; Based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, a dataset is generated; wherein, the operation data includes: no click, click, and purchase; The initial product ranking model is trained using the dataset to generate the pre-trained product ranking model.

3. The method according to claim 2, characterized in that, The method further includes: Obtain the actual interaction data of the product sorting results, wherein the actual interaction data includes at least one of the following: click data and purchase data; Based on the actual interaction data, determine whether the product ranking result is a high-quality ranking result, and obtain the judgment result; If the judgment result is that the product ranking result is a high-quality ranking result, then new training samples are generated based on the product information, user profile data, user behavior data, product access scenario data, and user operation data of each product in the high-quality ranking result corresponding to the high-quality ranking result. The pre-trained product ranking model is updated using the newly added training samples.

4. The method according to claim 2, characterized in that, The method further includes: Acquire multimodal data of the multiple products to be displayed; the multimodal data includes product text data, product image data, and product video data; The dataset is generated based on the product information of the multiple products to be displayed, the user profile data, user behavior data, product access scenario data of each sample user, and the user's operation data on each of the products to be displayed in the user behavior data, including: A dataset is generated based on the product information and multimodal data of the multiple products to be displayed, the user profile data of each sample user, the user behavior data, the product access scenario data, and the user's operation data on each of the products to be displayed in the user behavior data; Before the step of generating product ranking results based on the processing results using the pre-trained product ranking model, the method further includes: In response to the display command sent by the client, the multimodal data of the multiple products to be displayed is obtained, and the multimodal data is preprocessed to obtain preprocessed multimodal data; The preprocessed multimodal data is added to the processing result.

5. The method according to claim 1, characterized in that, After the step of generating product ranking results based on the processing results using a pre-trained product ranking model, the method further includes: Obtain at least one featured recommended product; Based on the at least one key recommended product and the product ranking result, a new product ranking result is generated.

6. The method according to any one of claims 1-5, characterized in that, After the steps of acquiring product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data, the method further includes: If the current user's user behavior data is not empty, then continue to execute the steps of preprocessing the product information of the multiple products to be displayed, the current user's user profile data, user behavior data, and product access scenario data to obtain the processing results, as well as subsequent steps; If the current user's user behavior data is empty, then a first cold start product ranking result is generated based on the popularity data of each of the products to be displayed; or Based on the user profile data of the current user and other users, calculate the similarity between the current user and other users; determine the closest user based on the similarity scores; and generate the first cold start product ranking result based on the user behavior data of the closest user. Based on the first cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

7. The method according to any one of claims 1-5, characterized in that, If the plurality of products to be displayed includes new products and old products, wherein products not existing in the current user's user behavior data are new products, and products existing in the current user's user behavior data are old products, then the step of generating product ranking results using a pre-trained product ranking model based on the processing results includes: The product information of the old products, the user profile data of the current user, the user behavior data, and the product access scenario data are processed and input into the pre-trained product ranking model to generate the product ranking result. The method further includes: Based on the product information of the new product and the old product, determine the product that is closest to the old product; The product sorting results are updated based on the sorting position of the closest old product to generate a second cold start product sorting result; Based on the second cold start product sorting result, the multiple products to be displayed are shown on the front-end page of the client.

8. A product display device, characterized in that, The device includes: The acquisition module is used to respond to the display command sent by the client and acquire product information of multiple products to be displayed, user profile data of the current user, user behavior data, and product access scenario data; wherein, the product access scenario data includes at least one of the following: client type and access time; The preprocessing module is used to preprocess the product information of the multiple products to be displayed, the user profile data of the current user, the user behavior data, and the product access scenario data to obtain the processing results. The sorting module is used to generate product sorting results based on the processing results using a pre-trained product sorting model. The display module is used to display the multiple products to be displayed on the front-end page of the client based on the product sorting results.

9. An electronic device, comprising: processor; as well as A memory configured to store computer-executable instructions, characterized in that, when executed, the executable instructions cause the processor to perform the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing one or more programs, characterized in that, When the one or more programs are executed by an electronic device including multiple applications, the electronic device causes the electronic device to perform the steps of the method as described in any one of claims 1-7.

Images