A product recommendation method and system based on big data analysis

Abstract

The present invention discloses a commodity recommendation method and system based on big data analysis, which relates to the technical field of intelligent recommendation systems, including collecting user behavior data on a platform and store data in the user's area; collecting public opinion data of commodities from social platforms and comment areas, integrating the behavior data, store data and public opinion data to form a recommendation data set; extracting behavior features from the behavior data, extracting store features from the store data, and extracting public opinion features from the public opinion data; integrating the behavior features, store features and public opinion features into a comprehensive feature vector, and generating a user portrait according to the comprehensive feature vector; constructing a commodity recommendation model based on the user portrait using a machine learning algorithm, taking the user portrait as input, and outputting a recommendation list of commodities; collecting real-time data of behavior data and store activities, dynamically adjusting the recommendation list, and recording user feedback on the recommended commodities.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent recommendation systems, and in particular to a product recommendation method and system based on big data analysis. Background Art

[0002] With the rapid development of e-commerce and online retail platforms, recommendation systems have become one of the key technologies for improving user experience and product sales. Traditional recommendation methods such as collaborative filtering and content-based recommendation algorithms can achieve recommendation functions to a certain extent, but when faced with large-scale, multi-dimensional data, there are problems of insufficient accuracy and poor flexibility. In recent years, with the help of big data and artificial intelligence technologies, recommendation systems have gradually begun to integrate user behavior data, product information, social public opinion data and geographic location-related data, trying to portray user portraits through multi-dimensional data to achieve more personalized recommendations. However, these technologies still face bottlenecks in practical applications, such as the complexity of multi-source data integration, insufficient real-time processing efficiency, and slow response to dynamic changes in user interests.

[0003] Although existing recommendation technologies have made some progress in improving the quality of recommendations, they still have obvious shortcomings. The analysis of user behavior data is mostly limited to simple click or purchase records, and lacks in-depth exploration of the temporal characteristics of behavior and interest dynamics; the use of store data relies more on static geographic location correlation and fails to fully incorporate the dynamic impact of store activities in the region; the processing of public opinion data often ignores the combined effect of emotional tendency intensity and public opinion heat. In addition, existing technologies are relatively simple in integrating multi-source heterogeneous data and mining interactive relationships, resulting in insufficient recommendation accuracy and personalization. At the same time, existing methods generally lack real-time update capabilities, making it difficult to dynamically adjust recommendation strategies based on changes in user behavior or store activities, and it is difficult to meet users' high demand for accurate recommendations. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] The present invention provides a product recommendation method based on big data analysis to solve the problem that the existing technology is relatively simple in integrating multi-source heterogeneous data and mining interactive relationships, resulting in insufficient recommendation accuracy and personalization.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a product recommendation method based on big data analysis, which comprises:

[0008] Collect user behavior data on the platform and store data in the user's area; collect product public opinion data from social platforms and comment areas, integrate behavior data, store data and public opinion data to form a recommendation data set and perform preprocessing;

[0009] Extract behavioral features from behavioral data, extract store features from store data, and extract public opinion features from public opinion data;

[0010] Integrate behavioral features, store features, and public opinion features into a comprehensive feature vector, and generate a user profile based on the comprehensive feature vector;

[0011] Based on user profiles, a product recommendation model is built using machine learning algorithms. The user profile is used as the model input. Combined with product sentiment and store activity data in the region, the product recommendation model outputs a list of recommended products.

[0012] Collect behavioral data and real-time data of store activities, dynamically adjust the recommendation list, and record user feedback on recommended products.

[0013] As a preferred solution of the product recommendation method based on big data analysis described in the present invention, wherein: collecting user behavior data on the platform and store data in the user's area; collecting product public opinion data from social platforms and comment areas, the specific steps are:

[0014] Combine the IP address and GPS location data of the user's device to initially obtain the user's geographic location;

[0015] Bind the user ID and geographic location to initially create records;

[0016] Obtain user behavior data on the platform through the product recommendation platform;

[0017] Based on the user's geographic location, filter the stores in the user's area from the store database;

[0018] Collect product-related public opinion data from social platforms and comment areas.

[0019] As a preferred solution of the product recommendation method based on big data analysis described in the present invention, the behavior data, store data and public opinion data are integrated to form a recommendation data set and pre-processed, specifically:

[0020] Integrate behavioral data, store data, and public opinion data to form a recommendation data set, which can be expressed as:

[0021] ;

[0022] in, Indicates the Recommendation dataset for users, Indicates the User behavior data, Indicates the Store data related to each user, Indicates the Public opinion data related to individual users;

[0023] Use a hash table to generate a unique hash value for each piece of data in the recommended dataset. Compare the hash values to see if they are duplicates. Duplicate data is deleted after hash collision detection.

[0024] Use the interquartile range (IQR) to detect outliers in the recommended data set, correct them, set an outlier threshold, and correct the data that exceeds the outlier threshold by the mean of similar data;

[0025] The K-nearest neighbor algorithm is used to predict the value of missing data based on the values of the k neighboring points that are most similar to the current data point, and then the missing data is completed using the KNN interpolation method.

[0026] As a preferred solution of the commodity recommendation method based on big data analysis described in the present invention, wherein: the extraction of behavioral features from behavioral data, the extraction of store features from store data, and the extraction of public opinion features from public opinion data are specifically performed as follows:

[0027] Perform aggregate analysis on behavioral data, count the frequency of users' behaviors within a certain time window, and extract the distribution characteristics of behavior types. The behavior feature extraction formula is:

[0028] ;

[0029] in, The feature vector representing the behavioral data, Indicates the time window length, 、 、 and Respectively represent the user at time Number of clicks, searches, browsing and purchases, Indicates the time period characteristics of user behavior;

[0030] Count the number of products on shelves and their distribution by category in the stores in the user's area, and extract the matching features between product categories and inventory. The formula for extracting store features is:

[0031] ;

[0032] in, The feature vector representing the store data, Indicates the number of stores in the user's area. The index variable representing the store, Indicates the The number of items on sale in each store, Indicates the The promotional offers of each store, Indicates that the user and The straight-line distance between stores, represents the distance weight function;

[0033] Count the number of times a product is mentioned on social media to extract its popularity features. Analyze the comments using a sentiment analysis model to extract the product's sentiment score. Combine the product's popularity and sentiment score to calculate the product's evaluation features. The formula for extracting public opinion features is:

[0034] ;

[0035] in, The feature vector representing the public opinion data, Indicates the total number of statistical public opinion data. Indicates the The number of forwarding of public opinion data, Indicates the Sentiment tendency score of a piece of public opinion data;

[0036] The steps of integrating behavioral features, store features, and public opinion features into a comprehensive feature vector and generating a user profile based on the comprehensive feature vector are as follows:

[0037] The eigenvector integration formula is:

[0038] ;

[0039] in, represents the integrated comprehensive feature vector;

[0040] According to the comprehensive feature vector Build a user profile for each user.

[0041] As a preferred solution of the product recommendation method based on big data analysis of the present invention, wherein: based on user portraits, a product recommendation model is constructed using a machine learning algorithm, and the specific steps are as follows:

[0042] The DeepFM model based on deep learning is used as the product recommendation model. It combines user portraits, product features, and dynamic features to output a user's interest prediction score for the product.

[0043] The product characteristics mainly refer to the basic information of the product and the public opinion heat characteristics. The calculation formula of the public opinion heat characteristics is:

[0044] ;

[0045] in, Indicates the product's public opinion heat score. The index variable representing the public opinion record, Indicates the The number of forwarding of public opinion data, Indicates the The sentiment tendency score of the public opinion data, represents the logarithmic function, Indicates the Number of comments on a piece of public opinion data;

[0046] The dynamic features refer to the real-time updates of store activity data and behavior data within a region. The regional store activity features can be calculated using the following formula:

[0047] ;

[0048] in, Indicates the activity intensity of stores in the area.

[0049] As a preferred solution of the product recommendation method based on big data analysis described in the present invention, the user portrait is used as the input of the model, and the product public opinion popularity and store activity data in the area are combined. The product recommendation model outputs a recommended list of products. The specific steps are as follows:

[0050] Map product features and dynamic features into a low-dimensional dense vector space to generate feature embedding representations;

[0051] For the feature cross-talk FM part of the product recommendation model, a factorization machine is used to capture the second-order interaction relationship between the input features to obtain the output of the FM part ;

[0052] The Deep part of the nonlinear feature learning of the product recommendation model is used to extract high-order nonlinear feature interactions using a fully connected deep neural network to obtain the output of the Deep part. ;

[0053] The outputs of the FM part and the Deep part are weighted and fused to obtain the final interest prediction score. , the expression is:

[0054] ;

[0055] in, Represents the Sigmoid function;

[0056] Predicting ratings based on interest , sort all products' interest prediction scores in descending order, and select the one with the highest score Generate a recommendation list of products and recommend a list Display to the user.

[0057] As a preferred solution of the product recommendation method based on big data analysis described in the present invention, the steps of collecting behavioral data and real-time data of store activities, dynamically adjusting the recommendation list, and recording user feedback on recommended products are as follows:

[0058] Collect user behavior data and feedback data in real time, and clean and pre-process them;

[0059] Dynamically calculate the recommendation priority score for each product based on user behavior data and feedback data , the expression is:

[0060] ;

[0061] in, Indicates that the user clicks on a product The total number of times, Indicates that the user purchased the product The total number of times, Indicates that the user is in the product Total time spent on the page, Indicates user's interest in the product Rating, represents the base of natural logarithms, and Represent the feature vectors of store data and public opinion data respectively, Indicates the time period characteristics of user behavior.

[0062] In a second aspect, the present invention provides a product recommendation system based on big data analysis, comprising an acquisition module, a feature extraction module, a product recommendation module, an adjustment feedback module, and a display module:

[0063] The acquisition module is responsible for collecting user-related data, pre-processing and integrating the data, and generating a recommendation data set;

[0064] The feature extraction module is responsible for extracting key features from the collected data and generating user profiles to provide input for the recommendation model;

[0065] The product recommendation module builds a recommendation model based on user portraits and product features, predicts user interest scores for products, and generates a recommendation list;

[0066] The adjustment feedback module is used to collect user behavior data and store activity data in real time, dynamically adjust the recommendation list, and record and analyze user feedback;

[0067] The display module is used to display the recommendation results to the user in a user-friendly manner and provide interactive functions.

[0068] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the product recommendation method based on big data analysis as described in the first aspect of the present invention is implemented.

[0069] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the product recommendation method based on big data analysis as described in the first aspect of the present invention.

[0070] The beneficial effects of the present invention are as follows: the present invention proposes a product recommendation method based on big data analysis through the process of data collection, preprocessing, feature extraction, model construction and dynamic adjustment. First, user behavior data, regional store data and product public opinion data are collected through the login module to construct a multi-dimensional recommendation data set to provide comprehensive data support for recommendation. Secondly, data accuracy and completeness are ensured through data cleaning, deduplication, anomaly correction and missing completion. Subsequently, user behavior characteristics, store activity characteristics and public opinion characteristics are extracted to generate a comprehensive feature vector for constructing user portraits. Based on the user portraits, a deep learning model is used to capture feature interactions, and a recommendation list is generated in combination with the dynamic characteristics of the product. The recommendation results are further optimized by real-time collection of user feedback and dynamic adjustment of recommendation priority. This method effectively solves the problems of poor data quality, delayed response and insufficient personalization in the existing technology, significantly improves the accuracy, real-time nature and user satisfaction of recommendations, and has important value in the field of intelligent recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0072] Figure 1 This is a flowchart of the product recommendation method based on big data analysis in Example 1.

[0073] Figure 2 This is a schematic diagram of generating a recommendation list in Example 1. DETAILED DESCRIPTION

[0074] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0075] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0076] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0077] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a product recommendation method based on big data analysis, comprising the following steps:

[0078] S1. Collect user behavior data on the platform and store data in the user's area; collect product public opinion data from social platforms and comment areas, integrate behavior data, store data and public opinion data to form a recommendation data set and perform preprocessing;

[0079] Combine the IP address and GPS location data of the user's device to initially obtain the user's geographic location;

[0080] Bind the user ID and geographic location to initially create records;

[0081] Based on the user's geographic location, filter the stores in the user's area from the store database, including store location, available products, and event information;

[0082] Obtain user behavior data on the product recommendation platform, including clicks, views, searches, and purchases, with each behavior accompanied by a timestamp and behavior type;

[0083] Behavior types, including clicks, views, searches, and purchases;

[0084] Based on the user's geographic location, filter the stores in the user's area from the store database, including store location, available products, and event information;

[0085] Collect product-related public opinion data from social platforms and comment areas, including user reviews, product popularity, and sentiment;

[0086] Preprocessing refers to deleting duplicate data, correcting abnormal data, and filling in missing data;

[0087] Integrate behavioral data, store data, and public opinion data to form a recommendation data set, which can be expressed as:

[0088] ;

[0089] in, Indicates the Recommendation dataset for users, Indicates the User behavior data, Indicates the Store data related to each user, Indicates the Public opinion data related to each user; a hash table is used to generate a unique hash value for each piece of data in the recommended data set, and the hash values are compared to see if they are repeated. Duplicate data is deleted after hash collision detection, thus achieving fast and efficient deduplication;

[0090] Delete duplicate records in behavioral data, store data, and public opinion data to ensure data uniqueness and accuracy;

[0091] Use the interquartile range (IQR) to detect outliers in the recommended data set, correct them, set an outlier threshold, and correct the data that exceeds the outlier threshold by the mean of similar data;

[0092] The threshold for outlier correction is set to IQR×1.5, and the expression is:

[0093] ;

[0094] in, represents the third quartile of the data set, Represents the first quartile of the data set;

[0095] For example, in the user stay time data, if = 60 seconds, =300 seconds, then data exceeding 300+1.5×(300-60)=660 seconds is considered an outlier and corrected by the mean of similar users;

[0096] Further explanation: IQR is an indicator used in statistics to measure the degree of dispersion of data, defined as the third quartile ( ) and the first quartile ( ), that is:

[0097] ;

[0098] IQR×1.5 is not used as the abnormal threshold directly, but 1.5 times the IQR is used to calculate the abnormal threshold. Extend upward to form a "tolerance range", and data points exceeding this range are judged as outliers, that is, the anomaly threshold is and 1.5 times the IQR.

[0099] The K-nearest neighbor algorithm is used to predict the value of missing data based on the values of the k neighboring points that are most similar to the current data point, and then the missing data is completed using the KNN interpolation method;

[0100] In the K-nearest neighbor algorithm for missing value completion, the k value is set to 5-15. For example, through cross-validation, it is found that when k=10, the interpolation accuracy (F1-score=0.92) is the highest and the computational efficiency is optimal;

[0101] The present invention constructs a multi-dimensional recommendation dataset through the steps of data integration and preprocessing, behavioral feature extraction, and introduction of public opinion data, and combines specific feature extraction methods and algorithm optimization to significantly improve the accuracy and timeliness of recommendations. The design of each step is based on the core requirements of recommendation. At the same time, by introducing technologies such as hash table deduplication, IQR outlier correction and KNN interpolation, the present invention is superior to existing technologies in data processing efficiency and recommendation effect. This technical solution not only solves the problem of user personalized recommendation, but also provides innovative ideas for the dynamic adjustment of recommendations.

[0102] S2. Extract behavioral features from behavioral data, store features from store data, and public opinion features from public opinion data;

[0103] Perform aggregate analysis on behavioral data, count the frequency of user behaviors within a certain time window, extract the distribution characteristics of behavior types, calculate the user's interest weights for different categories of goods, extract the preference characteristics of specific product categories, and extract the time characteristics of user behavior. The behavioral feature extraction formula is:

[0104] ;

[0105] in, The feature vector representing the behavioral data, Indicates the time window length, 、 、 and Respectively represent the user at time Number of clicks, searches, browsing and purchases, Indicates the time period characteristics of user behavior, with a value range of [0,1], indicating the distribution of user behavior activity within a day;

[0106] For example, user C's click, search, browse, and purchase times within a 7-day window are [45, 8, 30, 2], and his temporal activity level h(t) = 0.82 (proportion of nighttime activity). Then:

[0107] ;

[0108] Average daily frequency of clicks = ;

[0109] Average daily frequency of search behavior = ;

[0110] Average daily frequency of browsing behavior = ;

[0111] Average daily frequency of buying behavior = ;

[0112] Indicates the distribution of user activity within a day, calculated by counting the frequency of user behavior in specific time periods (such as early morning, morning, afternoon, and night).

[0113] ;

[0114] In behavioral data aggregation analysis, the time window length The value range is 7 to 30 days. For example, when When T = 7 days, it can capture users' recent high-frequency behaviors (such as click volume during promotions); when T = 30 days, it can cover users' medium- and long-term interest distribution (such as seasonal product preferences);

[0115] By independently counting clicks, searches, browsing, and purchases, we can fully reflect the intensity of user interest. For example, if users frequently search for a certain type of product but do not purchase it, it indicates that there is potential demand for that type of product.

[0116] Count the number of products on shelves and their distribution by category in stores in the user's area, extract matching features between product categories and inventory, extract store activity features (such as promotional discounts and limited-time offers), and calculate the intensity of these activities. Extract the distance features between the store's geographic location and the user, and measure the relevance between the store and the user using the distance weight. The formula for extracting store features is:

[0117] ;

[0118] in, The feature vector representing the store data, Indicates the number of stores in the user's area, The index variable representing the store, Indicates the The number of items on sale in each store, Indicates the The promotion intensity of each store is in the range of [0,1]. The higher the value, the more attractive the store promotion is. Indicates that the user and The straight-line distance between stores, Represents a distance weight function, which is used to calculate the weight of the distance effect on recommendations. The value range is (0, 1], with closer distances giving higher weights. This function uses the user's geographic location to filter product and event data from nearby stores to avoid recommending products that are beyond the user's reach. For example, products from stores that are closer and have stronger events are more attractive, improving the relevance of recommendations.

[0119] Count the number of times a product is mentioned on social media to extract its popularity features. Analyze the comments using a sentiment analysis model to extract the product's sentiment score. Combine the product's popularity and sentiment score to calculate the product's evaluation features. The formula for extracting public opinion features is:

[0120] ;

[0121] in, The feature vector representing the public opinion data, Indicates the total number of statistical public opinion data. Indicates the The number of forwarding of public opinion data, Indicates the The sentiment tendency score of the public opinion data (calculated by the sentiment analysis model, ranging from [0, 1]);

[0122] For example, there are 100 comments. Among the 100 comments, the total number of public opinion data is =100, average forwarding volume =350, sentiment score =0.85,

[0123] ;

[0124] Total forwarding volume ;

[0125] Assume that the number of forwarding of each public opinion is the same:

[0126] .

[0127] ;

[0128] Total sentiment score ;

[0129] Assume that the sentiment score of each public opinion is the same:

[0130] ;

[0131] ;

[0132] because =350 and =0.85:

[0133] ;

[0134] ;

[0135] ;

[0136] Public opinion features combine product popularity and sentiment, enabling recommendations to prioritize highly popular and highly rated products. For example, a product with high popularity and positive reviews on social media is more likely to attract clicks.

[0137] The present invention constructs a multi-dimensional feature vector model by extracting behavioral characteristics, store characteristics and public opinion characteristics. The design of each feature vector is based on user behavior, regional factors and market feedback, which solves the problems of insufficient personalization, lack of regionality and neglect of public opinion in traditional technologies. It innovatively introduces time-sensitive characteristics, distance weight functions and public opinion cross-features, which significantly improves the accuracy of recommendations and user experience.

[0138] S3. Integrate behavioral features, store features, and public opinion features into a comprehensive feature vector, and generate a user profile based on the comprehensive feature vector;

[0139] The eigenvector integration formula is:

[0140] ;

[0141] in, represents the integrated comprehensive feature vector;

[0142] Using PCA (Principal Component Analysis) Compressing from high dimensions (e.g., 200 dimensions) to 128 dimensions, retaining 95% of the original information and improving computational efficiency;

[0143] The covariance matrix is calculated based on the comprehensive eigenvector, and the expression is:

[0144] ;

[0145] in, represents the covariance matrix, represents the total number of samples, represents the index of the sample, Indicates the The characteristic values of the samples, Represents the mean vector of the data set;

[0146] Decompose the eigenvalue into:

[0147] ;

[0148] in, represents the eigenvector matrix, represents a diagonal matrix with dimensions , the diagonal elements are eigenvalues Sort by descending order , the off-diagonal elements are 0, represents the transpose of the eigenvector matrix;

[0149] Before selection The eigenvector (principal component) corresponding to the largest eigenvalue projects the data into a low-dimensional space, retaining the main information so that the cumulative variance contribution rate is ≥95%. The expression is:

[0150] ;

[0151] in, For the eigenvalues, For example, if the original dimension is 200 and the first 128 principal components are selected, the cumulative variance accounts for 95% and the inference speed is increased by 40%.

[0152] According to the comprehensive feature vector Build a user profile for each user;

[0153] Comprehensive feature vector The generation of recommendations effectively combines user behavior, geographical restrictions, and market feedback. For example, if a user frequently browses a certain type of product (behavioral characteristics), a nearby store is promoting the product (store characteristics), and the product has a high rating on social media (public opinion characteristics), the product can be recommended first, significantly improving the recommendation hit rate and user satisfaction.

[0154] The present invention uses behavioral characteristics , store characteristics and public opinion characteristics Integrated into comprehensive feature vector , a multi-dimensional user portrait generation method was constructed. The design of the comprehensive feature vector solved the problems of insufficient information dimension and strong data isolation in traditional recommendation methods, and significantly improved the comprehensiveness and dynamism of user portraits. In addition, by incorporating geographic location and market feedback into the user portrait generation process, the performance in regional recommendation and social recommendation was optimized. Finally, the present invention can generate multi-dimensional and high-precision user portraits in a shorter time, providing a more accurate and real-time means of expressing user needs for the recommendation method.

[0155] S4. Build a product recommendation model based on user profiles using machine learning algorithms. The model uses user profiles as input, combined with product sentiment and store activity data in the region, to output a list of recommended products.

[0156] The DeepFM model based on deep learning is used as the product recommendation model. It combines user portraits, product features, and dynamic features to output a user's interest prediction score for the product.

[0157] The product characteristics mainly refer to the basic information of the product (category, price and inventory) and the public opinion heat characteristics. The calculation formula of the public opinion heat characteristics is:

[0158] ;

[0159] in, Indicates the product's public opinion heat score. The index variable representing the public opinion record, Indicates the The number of forwarding of public opinion data, Indicates the The sentiment tendency score of the public opinion data, represents the logarithmic function, Indicates the Number of comments on a piece of public opinion data;

[0160] The threshold for public opinion heat is set at H ≥ 0.65. For example, when H ≥ 0.65, the click-through rate (CTR) of recommended products reaches 12.4%, which is a 143% increase compared to low-heat products (CTR = 5.1%).

[0161] CTR is calculated by counting users' clicks on recommended products. Impressions are the total number of times a product is exposed in the recommendation list, and clicks are the number of actual clicks. For example, if a product is displayed 1,000 times and clicked 124 times, the CTR is 12.4%, expressed as:

[0162] ;

[0163] The dynamic features refer to the real-time updates of store activity data and behavior data within a region. The regional store activity features can be calculated using the following formula:

[0164] ;

[0165] in, Indicates the intensity of store activity in the region;

[0166] The store distance weight function adopts an exponential decay model:

[0167] ;

[0168] For example, when =5km, ; =10km, , effectively reflecting the rule that "the closer the distance, the higher the recommendation priority";

[0169] The size of depends on the combined effect of store activity intensity and distance weight. When the store activity intensity is high and the distance is close, The value increases;

[0170] Map product features and dynamic features into a low-dimensional dense vector space to generate feature embedding representations;

[0171] For the feature cross-talk FM part of the product recommendation model, a factorization machine is used to capture the second-order interaction relationship between the input features to obtain the output of the FM part ;

[0172] The Deep part of the nonlinear feature learning of the product recommendation model is used to extract high-order nonlinear feature interactions using a fully connected deep neural network to obtain the output of the Deep part. ;

[0173] The outputs of the FM part and the Deep part are weighted and fused to obtain the final interest prediction score. , the expression is:

[0174] ;

[0175] in, Represents the Sigmoid function, which is used to normalize the output to the [0,1] interval;

[0176] Predicting ratings based on interest , sort all products' interest prediction scores in descending order, and select the one with the highest score Generate a recommendation list of products and recommend a list The recommendation list is displayed to the user and the formula for generating it is:

[0177] ;

[0178] in, Indicates a list of recommendations. Indicates the number of recommended products. The threshold for interest score (adjustable parameter, default value is 0.5), Indicates the Products, Represents the user, Represents a user For products Interest prediction score.

[0179] S5. Collect behavioral data and real-time data on store activities, dynamically adjust the recommendation list, and record user feedback on recommended products;

[0180] Collect user behavior data and feedback data in real time, and clean and pre-process it. Pre-processing includes deduplication and outlier processing.

[0181] Dynamically calculate the recommendation priority score for each product based on user behavior data and feedback data ,Adjust the recommendation sequence based on the recommendation priority score,recommendation priority score The expression is:

[0182] ;

[0183] in, Indicates that the user clicks on a product The total number of times, Indicates that the user purchased the product The total number of times, Indicates that the user is in the product Total time spent on the page, Indicates user's interest in the product The rating range is [0,1]. The higher the value, the higher the user's satisfaction with the product. represents the base of natural logarithms, and Represent the feature vectors of store data and public opinion data respectively, Indicates the time period characteristics of user behavior;

[0184] The weight distribution of user behavior time period features is:

[0185] 8:00-12:00: Weight 0.3 (for example, user browsing behavior is more dispersed);

[0186] 12:00-18:00: Weight 0.5 (for example, increased activity during lunch);

[0187] 18:00-22:00: Weight 0.8 (for example, the click-through rate in the evening is 3.2 times that in the early morning);

[0188] 22:00-8:00: Weight 0.1 (for example, user browsing behavior is more dispersed);

[0189] For example, using the DeepFM model, the comprehensive features of user C For example:

[0190] FM part: Calculate the second-order cross-features between product F and user C, with a weight of 0.32;

[0191] The expression of the second-order cross feature is:

[0192] ;

[0193] in, is the second-order cross feature, and is the feature index, is the feature dimension, , is the latent vector, and is the input eigenvalue, is the latent vector dot product.

[0194] Deep part: extracts nonlinear relationships through a 3-layer neural network (128→64→32), and outputs a value of 0.68.

[0195] Final Rating: (Exceeding the threshold of 0.5, added to the recommendation list);

[0196] During the dynamic adjustment phase, the formula for calculating the recommendation priority score of product G is: ,in:

[0197] , , ,

[0198] , , ,

[0199] , , ,

[0200] but ;

[0201] Recommendation priority score of product G ≈285.44, indicating that Product G will be given priority in dynamic adjustments due to its high public opinion popularity, strong activity intensity, and positive user feedback.

[0202] The present invention collects user behavior data and feedback data in real time and dynamically calculates the recommendation priority score. , and combined with behavioral characteristics, store characteristics and public opinion characteristics, it realizes the dynamic optimization of the recommendation method and the recommendation priority score The formula design comprehensively considers the depth of user interest (clicks, purchases, and dwell time), satisfaction (ratings), regionality (store characteristics), and market feedback (public opinion characteristics), and improves the real-time and accuracy of recommendation results through time-sensitivity adjustment. This solution not only solves the limitations of static and single-dimensional modeling of traditional recommendation methods, but also improves the adaptability and user satisfaction of recommendation methods through closed-loop optimization of user feedback.

[0203] This embodiment also provides a product recommendation system based on big data analysis, including: an acquisition module, a feature extraction module, a product recommendation module, an adjustment feedback module, and a display module:

[0204] The acquisition module is responsible for collecting user-related data, preprocessing and integrating the data, and generating a recommended data set;

[0205] The feature extraction module is responsible for extracting key features from the collected data and generating user profiles to provide input for the recommendation model;

[0206] The product recommendation module builds a recommendation model based on user portraits and product features, predicts user interest scores for products, and generates a recommendation list;

[0207] Adjust the feedback module to collect user behavior data and store activity data in real time, dynamically adjust the recommendation list, and record and analyze user feedback;

[0208] The display module is used to display the recommendation results to users in a user-friendly manner and provide interactive functions.

[0209] This embodiment also provides a computer device suitable for the case of a product recommendation method based on big data analysis, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the product recommendation method based on big data analysis proposed in the above embodiment.

[0210] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0211] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the product recommendation method based on big data analysis proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0212] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A product recommendation method based on big data analysis, characterized by: include: Collect user behavior data on the platform and store data in the user's area; Collect product public opinion data from social platforms and comment areas, integrate behavioral data, store data, and public opinion data to form a recommendation data set and perform preprocessing; Extract behavioral features from behavioral data, extract store features from store data, and extract public opinion features from public opinion data; Integrate behavioral features, store features, and public opinion features into a comprehensive feature vector, and generate a user profile based on the comprehensive feature vector; Based on user portraits, a product recommendation model is built using machine learning algorithms. The user portrait is used as the model input. Combined with the product's public opinion popularity and store activity data in the region, the product recommendation model outputs a recommended list of products. The specific steps are as follows: Map product features and dynamic features into a low-dimensional dense vector space to generate feature embedding representations; For the feature cross-talk FM part of the product recommendation model, a factorization machine is used to capture the second-order interaction relationship between the input features to obtain the output of the FM part ; The Deep part of the nonlinear feature learning of the product recommendation model is used to extract high-order nonlinear feature interactions using a fully connected deep neural network to obtain the output of the Deep part. ; The outputs of the FM part and the Deep part are weighted and fused to obtain the final interest prediction score; Collect behavioral data and real-time data on store activities, dynamically adjust the recommendation list, and record user feedback on recommended products. The specific steps are as follows: Collect user behavior data and feedback data in real time, and clean and pre-process them; Based on user behavior data and feedback data, the recommendation priority score of each product is dynamically calculated.

2. The product recommendation method based on big data analysis according to claim 1, characterized in that: The specific steps of collecting user behavior data on the platform and store data in the user's area and collecting product public opinion data from social platforms and comment areas are as follows: Combine the IP address and GPS location data of the user's device to initially obtain the user's geographic location; Bind the user ID and geographic location to initially create records; Obtain user behavior data on the platform through the product recommendation platform; Based on the user's geographic location, filter the stores in the user's area from the store database; Collect product-related public opinion data from social platforms and comment areas.

3. The product recommendation method based on big data analysis according to claim 2, characterized in that: The behavioral data, store data and public opinion data are integrated to form a recommendation data set and pre-processed, specifically: Integrate behavioral data, store data, and public opinion data to form a recommendation data set, which can be expressed as: ； in, Indicates the Recommendation dataset for users, Indicates the User behavior data, Indicates the Store data related to each user, Indicates the Public opinion data related to individual users; Use a hash table to generate a unique hash value for each piece of data in the recommended dataset. Compare the hash values to see if they are duplicates. Duplicate data is deleted after hash collision detection. Use the interquartile range (IQR) to detect outliers in the recommended data set, correct them, set an outlier threshold, and correct the data that exceeds the outlier threshold by the mean of similar data; Through the K-nearest neighbor algorithm, the value of the missing data is predicted based on the values of the k neighboring points most similar to the current data point, and then the missing data is completed through the KNN interpolation method.

4. The product recommendation method based on big data analysis according to claim 3, characterized in that: The specific steps of extracting behavioral features from behavioral data, extracting store features from store data, and extracting public opinion features from public opinion data are as follows: Perform aggregate analysis on behavioral data, count the frequency of users' behaviors within a certain time window, and extract the distribution characteristics of behavior types. The behavior feature extraction formula is: ； in, The feature vector representing the behavioral data, Indicates the time window length, 、 、 and Represents the user at time Number of clicks, searches, browsing and purchases, Indicates the time period characteristics of user behavior; Count the number of products on shelves and their distribution by category in the stores in the user's area, and extract the matching features between product categories and inventory. The formula for extracting store features is: ； in, The feature vector representing the store data, Indicates the number of stores in the user's area, The index variable representing the store, Indicates the The number of items on sale in each store, Indicates the The promotional offers of each store, Indicates that the user and The straight-line distance between stores, represents the distance weight function; Count the number of times a product is mentioned on social media to extract its popularity features. Analyze the comments using a sentiment analysis model to extract the product's sentiment score. Combine the product's popularity and sentiment score to calculate the product's evaluation features. The formula for extracting public opinion features is: ； in, The feature vector representing the public opinion data, Indicates the total number of statistical public opinion data. Indicates the The number of forwarding of public opinion data, Indicates the Sentiment tendency score of a piece of public opinion data; The steps of integrating behavioral features, store features, and public opinion features into a comprehensive feature vector and generating a user profile based on the comprehensive feature vector are as follows: The eigenvector integration formula is: ； in, represents the integrated comprehensive feature vector; According to the comprehensive feature vector Build a user profile for each user.

5. The product recommendation method based on big data analysis according to claim 4, characterized in that: The product recommendation model is constructed based on user portraits using machine learning algorithms. The specific steps are as follows: The DeepFM model based on deep learning is used as the product recommendation model. It combines user portraits, product features, and dynamic features to output a user's interest prediction score for the product. The product characteristics mainly refer to the basic information of the product and the public opinion heat characteristics. The calculation formula of the public opinion heat characteristics is: ； in, Indicates the product's public opinion heat score. The index variable representing the public opinion record, Indicates the The number of forwarding of public opinion data, Indicates the The sentiment tendency score of the public opinion data, represents the logarithmic function, Indicates the Number of comments on a piece of public opinion data; The dynamic features refer to the real-time updates of store activity data and behavior data within a region. The regional store activity features are calculated using the following formula: ； in, Indicates the activity intensity of stores in the area.

6. A product recommendation system based on big data analysis, based on the product recommendation method based on big data analysis according to any one of claims 1 to 5, characterized in that: Including acquisition module, feature extraction module, product recommendation module, adjustment feedback module and display module: The acquisition module is responsible for collecting user-related data, pre-processing and integrating the data, and generating a recommendation data set; The feature extraction module is responsible for extracting key features from the collected data and generating user profiles to provide input for the recommendation model; The product recommendation module builds a recommendation model based on user portraits and product features, predicts user interest scores for products, and generates a recommendation list; The adjustment feedback module is used to collect user behavior data and store activity data in real time, dynamically adjust the recommendation list, and record and analyze user feedback; The display module is used to display the recommendation results to the user in a user-friendly manner and provide interactive functions.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the product recommendation method based on big data analysis described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the product recommendation method based on big data analysis described in any one of claims 1 to 5 are implemented.

Images