E-commerce item selection method and system based on AI

Abstract

The invention discloses an AI-based e-commerce commodity selection method and system. The method comprises the steps of collecting multi-dimensional data of all commodity information uploaded by suppliers on a supply chain platform; preprocessing the multi-dimensional data to obtain preprocessed supplied commodity feature data; acquiring historical behavior data and current demand data of a purchaser on a supply chain platform to predict purchased commodity feature data preferred by the purchaser; and calculating similarity scores of the purchased commodity feature data and the supplied commodity feature data, and generating a personalized supplied commodity push list for the purchaser according to the similarity scores. By using the embodiment of the invention, the commodity information uploaded by the supplier and the historical behavior data of the purchaser can be comprehensively collected and preprocessed, the commodity characteristics capable of reflecting the preference of the purchaser can be effectively extracted, the commodity selection efficiency and accuracy of a supply chain platform are improved, and a better commodity recommendation service is provided for the purchaser.

Description

Technical Field

[0001] The present invention belongs to the technical field of data mining recommendation, and in particular to an e-commerce product selection method and system based on AI. Background Art

[0002] In the current context of rapid globalization and informatization, the importance of supply chain management has become increasingly prominent. The efficient operation of the supply chain not only depends on the close cooperation of all links, but also relies on a scientific and reasonable product selection mechanism to meet the ever-changing market demands and customer preferences. Traditional product selection methods often rely on experience judgment or simple market feedback, which makes it difficult to adapt to the changing market environment in real time, resulting in a series of problems such as low procurement efficiency, inventory backlogs, and reduced customer satisfaction. With the rise of e-commerce and online trading platforms, more and more companies are beginning to integrate their supply chain management into digital platforms. These platforms not only facilitate efficient docking between suppliers and buyers, but also generate massive amounts of commodity and user data. These data contain rich information, such as commodity categories, prices, sales, reviews, buyers' browsing and purchase records, etc. Multi-dimensional data provides a good foundation for optimizing product selection. However, relying solely on the accumulation of data is not enough to achieve intelligent and precise management of the supply chain. How to effectively process and analyze these data to achieve accurate push in a complex environment has become the main challenge facing the supply chain platform. Summary of the invention

[0003] The purpose of the present invention is to provide an AI-based e-commerce product selection method and system to address the deficiencies in the prior art, which can comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract product features that can reflect the preferences of buyers, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services.

[0004] An embodiment of the present application provides an AI-based e-commerce product selection method, the method comprising: Collect multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; Preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; Obtain buyers’ historical behavior data and current demand data on the supply chain platform to predict buyer’s preferred purchase commodity feature data; The similarity scores between the purchased commodity feature data and the feature data of each supply commodity are calculated, and a personalized supply commodity push list is generated for the purchaser based on the similarity scores.

[0005] Optionally, the multi-dimensional data includes at least: product description, price, sales volume, and evaluation data; The preprocessing of the multi-dimensional data to obtain preprocessed supply commodity feature data includes: Classifying the multi-dimensional data, wherein the classification types include at least numerical data and character data; Perform outlier processing on the numerical data, perform feature extraction on the numerical data and character data after outlier processing, and obtain numerical features and character features; The character features are converted into numerical feature data, and each numerical feature data is normalized and regularized so that features of different dimensions can be compared at the same scale.

[0006] Optionally, the acquisition of historical behavior data and current demand data of buyers on the supply chain platform to predict the characteristic data of purchased goods preferred by buyers includes: Obtain the historical supply commodities browsed or purchased by buyers from the historical behavior data on the supply chain platform, as well as the interaction data with the historical supply commodities; Calculating the buyer's interest score for each historically supplied commodity based on the interaction data; According to the current demand data, filter the historical supply products that meet one or more demand dimensions in the current demand data; Based on the selected historical supply commodities, the dimension features of the historical supply commodities corresponding to each demand dimension met are extracted; Based on the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity, the characteristics of the purchased commodities preferred by the buyers are constructed.

[0007] Optionally, the constructing of the purchase commodity characteristics preferred by the purchaser according to the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity includes: If there are multiple historical supply commodity dimension features of a demand dimension, the multiple historical supply commodity dimension features are weighted by using the interest scores of the multiple historical supply commodities to obtain a weighted dimension feature of the historical supply commodity; The weighted dimension features of historical supply commodities corresponding to each demand are combined to obtain the purchase commodity features preferred by the buyer.

[0008] Another embodiment of the present application provides an AI-based e-commerce product selection system, the system comprising: The collection module is used to collect multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; A processing module, used for preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; The prediction module is used to obtain the buyers’ historical behavior data and current demand data on the supply chain platform to predict the buyer’s preferred purchase commodity feature data; The push module is used to calculate the similarity score between the purchased commodity feature data and the feature data of each supply commodity, and generate a personalized supply commodity push list for the purchaser based on the similarity score.

[0009] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0010] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0011] Compared with the prior art, the present invention provides an AI-based e-commerce product selection method, which collects multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; pre-processes the multi-dimensional data to obtain pre-processed supply product feature data; obtains the historical behavior data and current demand data of buyers on the supply chain platform to predict the purchase product feature data preferred by the buyers; calculates the similarity score between the purchase product feature data and the feature data of each supply product, and generates a personalized supply product push list for the buyers based on the similarity score, so as to comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract the product features that can reflect the preferences of buyers, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A hardware structure block diagram of a computer terminal for an AI-based e-commerce product selection method provided in an embodiment of the present invention; Figure 2 A flowchart of an AI-based e-commerce product selection method provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of an AI-based e-commerce product selection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0014] The embodiment of the present invention first provides an AI-based e-commerce product selection method, which can be applied to electronic devices such as computer terminals, specifically ordinary computers, etc.

[0015] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for an AI-based e-commerce product selection method provided by an embodiment of the present invention. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the computer terminal may also include a transmission device 106 for communication functions and an input and output device 108. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0016] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / modules corresponding to the AI-based e-commerce product selection method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0017] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0018] See also Figure 2 , an embodiment of the present invention provides an e-commerce product selection method based on AI, which may include the following steps: S201, collects multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; Specifically, multi-dimensional data may include: product description, price, sales volume, evaluation data, etc. By fully acquiring all kinds of information about products, we can ensure the richness and diversity of data and provide a solid foundation for subsequent analysis and decision-making. One implementation method may include: 1. Determine the data source Determine the specific data source for extracting product information from the supply chain platform's database, including product descriptions, prices, sales, evaluation data, etc. These data sources should cover all uploaded products to ensure comprehensiveness and accuracy. 2. Data crawling mechanism design Design an automated data crawling system to regularly obtain new product information uploaded by suppliers through API interfaces or crawler technology. This system needs to set the crawling frequency and time period to ensure real-time performance and data freshness. 3. Multi-dimensional data structure definition Define the structure of multi-dimensional data, including but not limited to: - Product description: including product name, brand, model, specifications and other text information. - Price: records numerical data such as current price, historical price, and promotion information. - Sales volume: including real-time sales quantity, historical sales trends, etc. - Evaluation data: including user ratings, evaluation content, comment time, etc. 4. After capturing product information, the data preprocessing module first cleans the data to remove duplicate, erroneous, and missing data. The format of each type of data is standardized to ensure data consistency. For example, the product description is uniformly converted to lowercase letters and special characters are removed. 5. Multi-dimensional data labeling and classification Label and classify the captured data, use natural language processing (NLP) technology to analyze product descriptions, and extract keywords and topic information. Named entity recognition (NER) algorithms can be used to identify key information such as brand names and models, and store them in categories. 6. Data storage structure design Establish a multi-dimensional data storage model and use a relational database or NoSQL database to store product information. Design a suitable table structure so that data of different dimensions can be associated through foreign keys to ensure query efficiency and flexibility. 7. Data update and version control A data update mechanism is established to incrementally update the existing product information in each captured data. At the same time, in order to maintain the accuracy and historical traceability of the data, a version control strategy is implemented to record the timestamp and change information of each data update. 8. Data quality detection and monitoring Establish a data quality monitoring system to check the integrity, accuracy and consistency of data by setting reasonable thresholds. For example, for price and sales data, set appropriate reasonable ranges, and issue timely alarms and conduct manual review when abnormal data appears. 9. Multi-dimensional data fusion and feature construction After all data are collected and stored, data fusion is performed, and feature engineering technology is used to integrate data of different dimensions, extract important features and construct feature vectors, laying the foundation for subsequent recommendation algorithms. 10. Open Data Interface Finally, build a set of open data interfaces to allow other modules or services to access and call these multi-dimensional product data, so as to facilitate the subsequent data analysis, model training, user recommendation and other functions.

[0019] Through the above steps, multi-dimensional commodity information from the supply chain platform can be efficiently and accurately collected and processed. This process not only improves the quality and uniformity of the data, but also lays a solid foundation for subsequent operations such as buyer preference prediction and commodity recommendation. With the continuous accumulation and updating of data, the system can continuously learn the changes in buyers' preferences, thereby optimizing commodity push, improving user experience, and ultimately realizing intelligent management of the supply chain. In addition, by opening up data interfaces, relevant parties can carry out more flexible application development and create more commercial value.

[0020] S202, preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; By collecting and preprocessing multi-dimensional data of product information uploaded by all suppliers on the supply chain platform, valuable product feature data can be extracted from it, thereby generating a personalized supply product push list for buyers. Cleaning and converting the original data to improve data quality and facilitate subsequent analysis and modeling can not only improve the level of intelligent supply chain management, but also effectively reduce the time and cost of buyers looking for suitable products.

[0021] Specifically, the multi-dimensional data may be classified, wherein the classification types include at least numerical data and character data; Classify the data in the product information according to the nature of the information, for example, separate numerical data such as price and sales volume from character data such as product description and brand. Classification helps with subsequent feature extraction and processing, so that different types of data can be appropriately analyzed and used.

[0022] Perform outlier processing on the numerical data, perform feature extraction on the numerical data and character data after outlier processing, and obtain numerical features and character features; Detect and process values ​​in numerical data that are obviously deviated from the normal range, such as replacing them with reasonable values ​​through strong assignment, to avoid adverse effects on the overall analysis. Ensure the validity and accuracy of the data and improve the quality of recommendations.

[0023] Extract key features from the data after outlier processing. Numerical features include mean value, standard deviation, etc. Character features may include category labels, brands, etc. Feature extraction is the core of data analysis, and the extracted features directly affect the accuracy of push notifications.

[0024] The character features are converted into numerical feature data, and each numerical feature data is normalized and regularized so that features of different dimensions can be compared at the same scale.

[0025] By converting character data into numerical form through encoding and other methods, such as applying One-Hot Encoding and other technologies, character features can be effectively utilized. By standardizing numerical features (such as Z-score standardization) or normalizing (such as Min-Max scaling), all features are within the same numerical range. Normalization and regularization can eliminate the dimensional effects between different features and improve push performance.

[0026] Through the preprocessing of the above steps, we can ensure that the obtained supply product feature data is of high quality, thus laying a solid foundation for the subsequent generation of personalized push lists. This process can maximize the value of data and improve the intelligence and efficiency of the supply chain.

[0027] S203, obtaining historical behavior data and current demand data of buyers on the supply chain platform to predict characteristic data of purchased goods preferred by buyers; The steps of obtaining buyers’ historical behavior data and current demand data on the supply chain platform are aimed at gaining a deep understanding of buyers’ preferences and needs, and then providing them with personalized product recommendations. The core of this process is: 1. Data collection: By obtaining buyers’ historical behavior data (such as browsing history, purchase history, etc.) and current demand data (such as purchasing intentions, demand lists, etc.), it can fully reflect buyers’ purchasing habits and demand characteristics. 2. Preference prediction: Combining historical behavior with current demand helps build accurate preference models and identify the characteristics of products that buyers tend to choose. 3. Intelligent recommendation: By analyzing and mining the behavior patterns of buyers and generating personalized product push lists, it can effectively improve procurement efficiency and enhance buyers' stickiness and satisfaction with the platform. This approach enables the platform to provide more personalized and precise services in an increasingly competitive market, thereby improving the overall user experience.

[0028] Specifically, it is possible to obtain the historical supply commodities browsed or purchased by the buyers in the historical behavior data on the supply chain platform, as well as the interaction data with the historical supply commodities; Record all data of buyers browsing, purchasing or interacting with products on the platform, including product ID, number of views, order records, etc. This data is the basis of behavioral analysis. By analyzing this historical data, we can understand which products buyers are interested in, or what types of products they tend to purchase, thus providing basic data for preference analysis.

[0029] Calculating the buyer's interest score for each historically supplied commodity based on the interaction data; Using the collected interaction data, calculate the buyer's interest score for each historically supplied product. These scores can be calculated based on a weighted calculation of multiple behaviors (such as browsing, purchasing, and evaluation). The interest score is the key to understanding the buyer's preferences. It can quantify their attention to different products and lay a data foundation for subsequent personalized recommendations.

[0030] Specifically, a calculation formula for interest score can be:

[0031] This formula aims to integrate multi-dimensional data and quantitatively evaluate the interests of buyers in order to generate personalized product recommendations. By quantifying the scores of each behavior, we can more comprehensively understand the buyers' attention and preferences for a certain historical supply product, thereby achieving more accurate push in the recommendation system.

[0032] Among them, the is the interest score of buyer p for historical supply product s, It is the browsing behavior score of buyer p on historical supply product s, which indicates the quantification of the buyer's browsing frequency or duration on a certain product. A high browsing number indicates that the buyer has a high degree of attention to the product. Used to smooth browsing behavior. Smoothing is done to reduce the impact of extreme browsing behavior and prevent high-frequency browsing data from causing excessive deviations in interest scores. This process gradually reduces the score increase brought about by increased browsing behavior, which helps to avoid over-amplifying the impact of the number of browsing times. Score the purchase behavior of buyer p for historically supplied product s, which directly reflects the buyer's purchase intention for the product. Purchase behavior is usually the strongest indicator for evaluating the attractiveness of a product. Used to increase the influence of purchase behavior, indicating that purchase behavior can exponentially increase the interest score. This shows that the intensity of purchase intention is very important for predicting the buyer's interest, emphasizing the decisive role of purchase behavior. The evaluation behavior score of the buyer p on the historical supply product s reflects the buyer's feedback and evaluation of the product. A high evaluation usually indicates that the quality and service of the product are recognized and may affect the purchase decisions of other buyers. The evaluation score directly affects the interest score, showing the buyer's trust and satisfaction with the product, which is an additional recognition of the value of the product. Score the interaction time of buyer p on historically supplied product s, which is a quantitative measure of the interaction time between the buyer and the product, and usually represents the potential degree of attention and "immersion" in the product. A longer interaction time usually means that the buyer is more interested in the product. The processing can ensure that the impact on the interaction time is non-linear, prevent excessively long interaction time from having a disproportionate impact on the score, and more reasonably reflect its contribution to the interest score. is the corresponding weight coefficient, which is used to adjust the relative importance of each behavior score in the total interest score. Different buyers may attach different importance to behaviors such as browsing, purchasing, evaluation and interaction time, so the setting of these weights can be adjusted according to actual conditions to optimize the recommendation effect.

[0033] This formula builds a comprehensive interest scoring system by integrating a large amount of historical behavior data, which helps to accurately explore and understand the needs and preferences of buyers. This provides a solid foundation for subsequent personalized recommendations and improves the overall user experience and satisfaction.

[0034] According to the current demand data, filter the historical supply products that meet one or more demand dimensions in the current demand data; Analyze the current demand data of buyers, match the characteristic dimensions of the goods they want to purchase (such as category, quantity, price range, etc.), and filter out historical supply goods that meet these needs. For example, if product a meets one demand dimension, product b meets two demand dimensions, and product c does not meet any current demand dimension, then a and b are filtered out. This step ensures that the recommended goods not only meet the buyers' historical interests, but also meet their current specific needs, improving the accuracy of the recommendation.

[0035] Based on the selected historical supply commodities, the dimension features of the historical supply commodities corresponding to each demand dimension met are extracted; For the selected historical supply products, extract their features under the current demand dimension, such as price, brand, specifications, etc. For example, extract the product dimension features of the above one demand dimension satisfied by product a, and extract the two product dimension features of the above two demand dimensions satisfied by product b. A product dimension feature is a partial feature of the entire product feature, and a partial feature corresponds to a demand dimension. By extracting and analyzing these features, we can further understand which product features meet the needs of buyers, so as to accurately locate suitable recommended products.

[0036] Based on the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity, the characteristics of the purchased commodities preferred by the buyers are constructed.

[0037] The dimensional features of each historically supplied product are combined with the corresponding interest score to form a high-dimensional feature vector. This feature vector contains the numerical performance of each product in different feature dimensions and interest score information. The constructed feature vector is used to subsequently calculate the similarity score between the buyer and the supplied product, which ensures that the recommendation system can identify and recommend the most relevant products based on the buyer's preferences, thereby improving the accuracy and effectiveness of the recommendation.

[0038] Specifically, if there are multiple historical supply commodity dimension features of a demand dimension, the multiple historical supply commodity dimension features are weighted by using the interest scores of the multiple historical supply commodities to obtain a weighted dimension feature of the historical supply commodity; Under a specific demand dimension, there may be multiple dimensional features of historically supplied goods that can be considered. The interest scores of these goods quantify the different levels of attention and preference of buyers for these goods. By weighting the dimensional features of these historically supplied goods and the corresponding interest scores, a comprehensive weighted dimensional feature can be obtained. This weighting is usually achieved by multiplying each product dimensional feature by the normalized interest score corresponding to the product to which it belongs, and then summing them up so that one demand dimension corresponds to only one weighted dimensional feature.

[0039] Through this weighting, the characteristics of historical supply products that are more important to buyers can be effectively highlighted, so that the buyers' real preferences can be more accurately reflected. This process helps to improve the personalization of the recommendation system, so that the recommended products can better meet the needs of buyers.

[0040] The weighted dimension features of historical supply commodities corresponding to each demand are combined to obtain the purchase commodity features preferred by the buyer.

[0041] After calculating the weighted dimension features under each demand dimension, these features need to be merged to form a unified feature vector as the purchase product feature that meets the buyer's preferences. This purchase product feature is the most ideal product feature constructed by comprehensive prediction, and the corresponding product can be understood as the ideal purchase product in the buyer's mind. This usually includes simple feature concatenation, or constructing the final feature representation through some aggregation function (such as average). The merged feature vector will contain the comprehensive preference information of the buyer on all demand dimensions.

[0042] The merging operation enables the model to comprehensively consider the characteristics of multiple demand dimensions and form a multi-dimensional comprehensive feature. This comprehensive feature can fully reflect the preferences of buyers, which helps the subsequent recommendation algorithm to make more effective recommendation decisions based on comprehensive information when calculating similarities, and ultimately improve the shopping experience of buyers on the platform.

[0043] S204, calculating a similarity score between the characteristic data of the purchased commodity and the characteristic data of each supply commodity, and generating a personalized supply commodity push list for the purchaser according to the similarity score.

[0044] On the supply chain selection platform, the buyer's personalized push list is calculated based on the similarity between the characteristics of the purchased goods and the characteristics of each supply product. By calculating the similarity score, the goods that meet the buyer's interests can be effectively identified. This similarity-based push algorithm not only improves the accuracy of the recommendation, but also effectively improves the user experience. Function and significance: 1. Accurate matching: The calculation of similarity scores can ensure that the pushed products meet the specific needs and preferences of buyers, improving the relevance of recommendations. 2. Dynamic adjustment: As buyer behavior data is continuously updated, the algorithm can adjust similarity scores in real time, thereby maintaining the effectiveness and timeliness of push notifications in a rapidly changing market environment. 3. Improve user satisfaction: Through accurate product matching, the satisfaction and purchase willingness of buyers are improved, thereby promoting the transaction volume of the platform. One implementation method may include: 1. Feature standardization: Before calculating the similarity score, the purchased goods feature data and all supplied goods feature data are standardized, such as using the Z-score standardization method, etc. This step ensures that different features are compared on the same dimension, laying the foundation for subsequent similarity calculations. 2. Similarity calculation method selection: You can choose cosine similarity as a similarity score calculation method. This method helps measure the directional similarity between two feature vectors regardless of their size. 3. Threshold filtering: Set a similarity threshold and only keep the supply products corresponding to the supply product characteristics above this threshold as the recommended objects. This step can avoid pushing irrelevant products to buyers and improve the quality of recommendations. 5. Personalized push list generation: Generate a personalized supply product push list through threshold filtering. This list can be directly displayed to buyers, including basic information and similarity scores of each recommended product to help buyers make purchasing decisions. Through the comprehensive implementation of the above steps, similarity scores can be effectively calculated, and personalized product recommendation lists can be provided to buyers based on these scores, thereby improving the practicality of the system and user satisfaction.

[0045] It can be seen that the multi-dimensional data of all product information uploaded by suppliers on the supply chain platform is collected; the multi-dimensional data is pre-processed to obtain the pre-processed supply product feature data; the historical behavior data and current demand data of the buyers on the supply chain platform are obtained to predict the purchase product feature data preferred by the buyers; the similarity score between the purchase product feature data and the feature data of each supply product is calculated, and a personalized supply product push list is generated for the buyers based on the similarity score, so as to comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract the product features that can reflect the buyers' preferences, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services.

[0046] Another embodiment of the present invention provides an AI-based e-commerce product selection system, see Figure 3 , the system may include: The collection module 301 is used to collect multi-dimensional data of all commodity information uploaded by suppliers on the supply chain platform; A processing module 302 is used to pre-process the multi-dimensional data to obtain pre-processed supply commodity feature data; Prediction module 303, used to obtain the historical behavior data and current demand data of buyers on the supply chain platform to predict the characteristic data of purchased goods preferred by buyers; The push module 304 is used to calculate the similarity score between the purchased commodity feature data and the feature data of each supply commodity, and generate a personalized supply commodity push list for the purchaser according to the similarity score.

[0047] It can be seen that the multi-dimensional data of all product information uploaded by suppliers on the supply chain platform is collected; the multi-dimensional data is pre-processed to obtain the pre-processed supply product feature data; the historical behavior data and current demand data of the buyers on the supply chain platform are obtained to predict the purchase product feature data preferred by the buyers; the similarity score between the purchase product feature data and the feature data of each supply product is calculated, and a personalized supply product push list is generated for the buyers based on the similarity score, so as to comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract the product features that can reflect the buyers' preferences, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services.

[0048] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0049] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps: S201, collects multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; S202, preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; S203, obtaining historical behavior data and current demand data of buyers on the supply chain platform to predict characteristic data of purchased goods preferred by buyers; S204, calculating a similarity score between the characteristic data of the purchased commodity and the characteristic data of each supply commodity, and generating a personalized supply commodity push list for the purchaser according to the similarity score.

[0050] It can be seen that the multi-dimensional data of all product information uploaded by suppliers on the supply chain platform is collected; the multi-dimensional data is pre-processed to obtain the pre-processed supply product feature data; the historical behavior data and current demand data of the buyers on the supply chain platform are obtained to predict the purchase product feature data preferred by the buyers; the similarity score between the purchase product feature data and the feature data of each supply product is calculated, and a personalized supply product push list is generated for the buyers based on the similarity score, so as to comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract the product features that can reflect the buyers' preferences, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services.

[0051] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0052] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0053] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program: S201, collects multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; S202, preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; S203, obtaining historical behavior data and current demand data of buyers on the supply chain platform to predict characteristic data of purchased goods preferred by buyers; S204, calculating a similarity score between the characteristic data of the purchased commodity and the characteristic data of each supply commodity, and generating a personalized supply commodity push list for the purchaser according to the similarity score.

[0054] It can be seen that the multi-dimensional data of all product information uploaded by suppliers on the supply chain platform is collected; the multi-dimensional data is pre-processed to obtain the pre-processed supply product feature data; the historical behavior data and current demand data of the buyers on the supply chain platform are obtained to predict the purchase product feature data preferred by the buyers; the similarity score between the purchase product feature data and the feature data of each supply product is calculated, and a personalized supply product push list is generated for the buyers based on the similarity score, so as to comprehensively collect and pre-process the product information uploaded by suppliers and the historical behavior data of buyers, effectively extract the product features that can reflect the buyers' preferences, improve the product selection efficiency and accuracy of the supply chain platform, and provide buyers with better product recommendation services.

[0055] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. An AI-based e-commerce product selection method, characterized in that: The method comprises: Collect multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; Preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; Obtain buyers’ historical behavior data and current demand data on the supply chain platform to predict buyer’s preferred purchase commodity feature data; The similarity scores between the purchased commodity feature data and the feature data of each supply commodity are calculated, and a personalized supply commodity push list is generated for the purchaser based on the similarity scores.

2. The method according to claim 1, characterized in that The multi-dimensional data includes at least: product description, price, sales volume, and evaluation data; The preprocessing of the multi-dimensional data to obtain preprocessed supply commodity feature data includes: Classifying the multi-dimensional data, wherein the classification types include at least numerical data and character data; Perform outlier processing on the numerical data, perform feature extraction on the numerical data and character data after outlier processing, and obtain numerical features and character features; The character features are converted into numerical feature data, and each numerical feature data is normalized and regularized so that features of different dimensions can be compared at the same scale.

3. The method according to claim 2, characterized in that The acquisition of the buyer's historical behavior data and current demand data on the supply chain platform to predict the buyer's preferred purchase commodity feature data includes: Obtain the historical supply commodities browsed or purchased by buyers from the historical behavior data on the supply chain platform, as well as the interaction data with the historical supply commodities; Calculating the buyer's interest score for each historically supplied commodity based on the interaction data; According to the current demand data, filter the historical supply products that meet one or more demand dimensions in the current demand data; Based on the selected historical supply commodities, the dimension features of the historical supply commodities corresponding to each demand dimension met are extracted; Based on the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity, the characteristics of the purchased commodities preferred by the buyers are constructed.

4. The method according to claim 3, characterized in that The characteristics of the purchased goods preferred by the buyers are constructed according to the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity, including: If there are multiple historical supply commodity dimension features of a demand dimension, the multiple historical supply commodity dimension features are weighted by using the interest scores of the multiple historical supply commodities to obtain a weighted dimension feature of the historical supply commodity; The weighted dimension features of historical supply commodities corresponding to each demand are combined to obtain the purchase commodity features preferred by the buyer.

5. An AI-based e-commerce product selection system, characterized in that: The system comprises: The collection module is used to collect multi-dimensional data of all product information uploaded by suppliers on the supply chain platform; A processing module, used for preprocessing the multi-dimensional data to obtain preprocessed supply commodity feature data; The prediction module is used to obtain the buyers’ historical behavior data and current demand data on the supply chain platform to predict the buyer’s preferred purchase commodity feature data; The push module is used to calculate the similarity score between the purchased commodity feature data and the feature data of each supply commodity, and generate a personalized supply commodity push list for the purchaser based on the similarity score.

6. The system according to claim 5, characterized in that The multi-dimensional data at least includes: product description, price, sales volume, and evaluation data; the processing module is specifically used to: Classifying the multi-dimensional data, wherein the classification types include at least numerical data and character data; Perform outlier processing on the numerical data, perform feature extraction on the numerical data and character data after outlier processing, and obtain numerical features and character features; The character features are converted into numerical feature data, and each numerical feature data is normalized and regularized so that features of different dimensions can be compared at the same scale.

7. The system according to claim 6, characterized in that The prediction module is specifically used for: Obtain the historical supply commodities browsed or purchased by buyers from the historical behavior data on the supply chain platform, as well as the interaction data with the historical supply commodities; Calculating the buyer's interest score for each historically supplied commodity based on the interaction data; According to the current demand data, filter the historical supply products that meet one or more demand dimensions in the current demand data; Based on the selected historical supply commodities, the dimension features of the historical supply commodities corresponding to each demand dimension met are extracted; Based on the dimensional characteristics of each historically supplied commodity and the interest score of each historically supplied commodity, the characteristics of the purchased commodities preferred by the buyers are constructed.

8. The system according to claim 7, characterized in that The prediction module is specifically used for: If there are multiple historical supply commodity dimension features of a demand dimension, the multiple historical supply commodity dimension features are weighted by using the interest scores of the multiple historical supply commodities to obtain a weighted dimension feature of the historical supply commodity; The weighted dimension features of historical supply commodities corresponding to each demand are combined to obtain the purchase commodity features preferred by the buyer.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images