Live broadcast e-commerce product promotion method and system

Abstract

The invention discloses a live broadcast e-commerce product promotion method, and relates to the technical field of live broadcast e-commerce, and the method comprises the steps: calculating the popularity of a product through a TF-IDF method based on the combination of an influence maximization algorithm and a collaborative filtering technology, quantifying the scarcity of the product and the preference of a user, reflecting the real preference of the user for the product, and improving the user experience. And through user multi-dimensional feature similarity, inaccurate recommendation when user data is sparse is avoided, a product promotion user group is refined and a user personalized product promotion strategy is designed in combination with the calculated user with the best propagation capability, and meanwhile, product promotion strategy implementation data is collected and analyzed, so that the recommendation accuracy of the product promotion strategy is improved. Market demand prediction, cost change analysis and a particle swarm optimization algorithm are combined with a dynamic adjustment pricing strategy, and real-time product pricing optimization is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of live e-commerce, and in particular to a method and system for promoting live e-commerce products. Background Art

[0002] With the rapid development of Internet technology, live e-commerce, as a new e-commerce model, has received extensive attention in recent years and has quickly become an important part of online retail. By combining video live streaming and online sales, live e-commerce realizes the real-time display and interactive promotion of products, enabling consumers to interact with the anchor and other consumers by watching the live stream, obtaining product purchase information, product usage experience, and promotional offers, thereby stimulating consumer demand. In addition, with the development of social media platforms, e-commerce platforms have also begun to rely on user behavior data on social networks, such as interactive data like likes, comments, and shares, combined with purchase history on the e-commerce platform, to enable merchants to accurately analyze user interests and conduct targeted personalized promotions.

[0003] Although the existing technology has been able to collect and analyze some user data through e-commerce platforms and social media, and has improved the promotion of live e-commerce products to a certain extent, the existing methods for promoting live e-commerce products still have many limitations, mainly reflected in several aspects. First, the existing analysis of user information relies on the user's basic information and purchase history. Traditional methods based on a single data source often suffer from incomplete data and inaccurate recommendations. Although some platforms have improved the accuracy of user portraits by introducing multi-dimensional data such as social media data and live streaming behavior data, it is still difficult to comprehensively evaluate the social influence and activity of users on multiple platforms. Second, when dividing the promotion groups in the existing technology, it often relies on static purchase behavior data and traditional recommendation algorithms, which makes the personalization degree of the recommendation system not high, difficult to respond promptly to the dynamic changes of user behavior, and thus affects the promotion effect. At the same time, most product pricing strategies rely on fixed pricing models and fail to be flexibly adjusted according to changes in user interest, product popularity, and market demand, thus missing opportunities to optimize sales. Therefore, with the intensification of market competition, how to accurately push personalized products to target users, flexibly adjust product pricing, and improve the purchase conversion rate of users has become an urgent problem for live e-commerce platforms. Summary of the Invention

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

[0005] Therefore, the present invention provides a method and system for promoting live e-commerce products, which solves the problems of the existing live e-commerce promotion methods in accurately pushing personalized products to target users and flexibly adjusting product pricing.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for promoting live e-commerce products, which includes,

[0008] Collecting key data of users through multiple channels and performing preprocessing, and dividing the product promotion groups by using keyword search and social influence evaluation on the processed data;

[0009] Based on the divided promotion groups, combining the influence maximization algorithm and collaborative filtering technology to optimize the promotion groups, and formulating personalized product promotion strategies;

[0010] When implementing product promotion, by collecting and analyzing the real-time behavior data of users, using the particle swarm optimization algorithm combined with multivariate regression analysis and time series analysis to predict future market demand and cost changes, and adjusting the product price in real time.

[0011] As a preferred solution of the method for promoting live e-commerce products according to the present invention, wherein: the dividing of the product promotion groups by using keyword search and social influence evaluation on the processed data means extracting the key features of users from the cleaned data, including the interest features of users , purchase habit features and social influence ;

[0012] And extracting keywords of the social interaction data of users through the TF-IDF method, and evaluating the weight of each keyword in the user interaction text ;

[0013] Extracting keywords for each user in the social platform and aggregating them to obtain the keyword vector of the user ;

[0014] Normalizing the interest features , purchase habit features , social influence and keyword vector and combining them by using weighted summation, and dividing the users into product promotion groups T according to similar features through the K-means clustering algorithm.

[0015] As a preferred solution of the method for promoting live e-commerce products according to the present invention, wherein: the optimizing the promotion groups by combining the influence maximization algorithm and collaborative filtering technology based on the divided promotion groups includes:

[0016] Converting the combined features into elements in the rating matrix, and constructing the rating matrix , with the behavior user u as the row and the product i as the column;

[0017] Collect the frequency of product i in the ratings of each user and calculate the TF-IDF weight based on product popularity ;

[0018] After calculating the TF-IDF weight of each product, correct the rating of product i to obtain the corrected rating ;

[0019] And collect user characteristics, including the age, occupation, and gender of the user, and calculate the similarity of user characteristics based on fuzzy logic;

[0020] Use the weighted average method to fuse the age, occupation, and gender similarities into a comprehensive similarity ;

[0021] Construct a cognitive behavior propagation graph based on the propagation relationship between users and products, where nodes represent users and products, and edges represent the behavior of users towards products, and calculate the propagation influence ;

[0022] Based on the propagation influence, use the influence maximization algorithm based on the greedy algorithm to find the users with the greatest influence on the behavior of other users, through the propagation influence Select the users with the greatest propagation influence as the initial seed user set S. In each round of iteration, select the user u that can maximize the influence propagation currently;

[0023] Based on the propagation influence, calculate the influence increment of each user, and sort the users in descending order according to the influence increment. Select the user with the greatest influence increment in the social network as the propagation node for propagation, and in each round of optimization, add users in descending order to increase the influence in the propagation network until the user influence increment is less than the preset threshold , then the contribution of the propagation influence to the final effect is the smallest, and the iterative optimization process stops;

[0024] By calculating the TF-IDF rating similarity and the propagation influence similarity, and perform a weighted sum with the user characteristic similarity ,

[0025] Use the K-Means++ algorithm to cluster the comprehensive user similarity and reassign a group to each user , remove the users in the assigned groups from all users, and perform clustering again according to the remaining user groups until all users are completely clustered and classified;

[0026] Set the comprehensive user similarity The threshold is b. When there is a comprehensive user similarity within the group greater than or equal to b, then the users in the group are the core user group. If the comprehensive user similarity within the group is less than b, then the users in the group are the potential user group.

[0027] As a preferred embodiment of the method for promoting live e-commerce products according to the present invention, wherein: the formulation of personalized product promotion strategies refers to formulating promotion strategies for the core user group and the potential user group respectively. For the core user group, set membership pricing and differential pricing, use the influence of the core user group for social sharing rewards and referral reward mechanisms to encourage core users to promote products, and launch bundle sales and package discounts. For the potential user group, by providing time-limited offers and group-buying discounts, encourage users to purchase and share products, and rewards can be obtained for successful promotion and purchase, and use the integral and cashback system to motivate users to purchase and participate in social sharing.

[0028] As a preferred embodiment of the method for promoting live e-commerce products according to the present invention, wherein: when implementing product promotion, collecting and analyzing real-time user behavior data for promotional strategy adjustment means labeling the core user group as group A and the potential user group as group B, respectively collecting the purchase data of users when the strategies of groups A and B are implemented, and calculating the purchase conversion rate of group A , calculating the purchase conversion rate of group B ;

[0029] When then the conversion rate of group B is higher than that of group A, and it is necessary to adjust the promotional strategy of group A, increase the discount rate of group A and launch time-limited offers. When , adjust the promotional strategy of group B, extend the preferential time and increase the discount rate, and encourage users in group B to purchase through group-buying.

[0030] As a preferred embodiment of the method for promoting live e-commerce products according to the present invention, wherein: using the particle swarm optimization algorithm combined with multivariate regression analysis and time series analysis to predict future market demand and cost changes, and real-time adjusting product pricing means using time series analysis to predict the market demand ;

[0031] using multivariate regression analysis to predict the cost changes of products in the future for a period of time, including production costs, logistics costs, warehousing costs, etc., and adjusting the pricing strategy;

[0032] Optimize multi-objective pricing by combining the particle swarm optimization and the sparrow search algorithm for user behavior, future market demand, and cost changes , and adjust the product pricing and promotion intensity in real time;

[0033] And according to the dynamically adjusted price Continuously optimize the product pricing using the particle swarm optimization algorithm. Each particle represents a pricing plan, and the goal of the particle is to optimize the pricing strategy. The particle continuously updates its velocity according to the current market environment and user behavior and position .

[0034] As a preferred solution of the live e-commerce product promotion method described in the present invention, wherein: the collecting and preprocessing of the key data of users through multiple channels refers to extracting the corresponding historical purchase records of users from the e-commerce platform database using SQL queries, regularly obtaining the interaction data of users on the social platform through web crawler technology from the user's social media, using the API of the integrated live platform to regularly obtain the behavior data of users during the live broadcast. After the data is collected, noise data is removed, missing values are filled, and data normalization is performed on the data.

[0035] In a second aspect, the present invention provides a live e-commerce product promotion system, including:

[0036] A data collection module, used to obtain user data from the e-commerce platform, social media, and live API, and perform noise removal, missing value filling, and data normalization;

[0037] A group division module, used to extract keywords of the interests, purchase habits, social influence, and social interaction data of users, and perform processing and merging, and perform clustering to divide the product promotion groups;

[0038] A group optimization module, used to collect user characteristics using the product promotion groups, calculate the similarity between users, and optimize the promotion groups through the greedy algorithm and collaborative filtering technology;

[0039] A strategy formulation module, used to provide exclusive pricing and referral rewards for core users, and provide time-limited discounts, group purchase discounts, and point rewards for socially active users;

[0040] A dynamic pricing module, used to predict market demand and cost, adjust the pricing through the particle swarm optimization algorithm, and optimize the sales conversion rate in real time.

[0041] In a third aspect, the present invention provides a computer device, including a memory and a processor, wherein: when the computer program is executed by the processor, any step of the live e-commerce product promotion method described in the first aspect of the present invention is implemented.

[0042] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by a processor, any step of the live e-commerce product promotion method described in the first aspect of the present invention is implemented.

[0043] The beneficial effects of the present invention are as follows: by combining the influence maximization algorithm and collaborative filtering technology, calculating the product popularity using the TF-IDF method, quantifying the product scarcity and user preferences, reflecting the true preferences of users for products, and through the user multi-dimensional feature similarity, avoiding inaccurate recommendations when user data is sparse. By combining the calculated users with the most dissemination ability, refining the product promotion user group, designing personalized product promotion strategies for users, and at the same time collecting and analyzing the data of the implementation of the product promotion strategy, using market demand prediction, cost change analysis and particle swarm optimization algorithm to dynamically adjust the pricing strategy, the real-time optimization of product pricing is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0045] Figure 1 It is a flowchart of the live e-commerce product promotion method in Embodiment 1.

[0046] Figure 2 It is a structural diagram of the live e-commerce product promotion system in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention with reference to the drawings in the specification.

[0048] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0049] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not all refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments.

[0050] Example 1, referring to Figure 1 and Figure 2 , which is the first embodiment of the present invention. This embodiment provides a method for promoting live e-commerce products, including the following steps:

[0051] S1. Collect key data of users through multiple channels and preprocess it. Use keyword search and social influence assessment on the processed data to divide the product promotion group;

[0052] Specifically, use SQL query to extract corresponding user historical purchase records from the e-commerce platform database, including the product categories purchased, purchase frequency, and preferences. Regularly obtain user interaction data on social platforms (such as Weibo, Instagram, Douyin, etc.) through web crawling technology, including social behaviors such as likes, comments, and shares. These data help evaluate the influence of users in social platforms and correlate with their purchase behaviors. Use the API of the integrated live platform to regularly obtain user behavior data during the live broadcast, including viewing duration, number of comments, number of likes, number of shares, etc. Through these data, the activity of users in the live broadcast and their interest in specific products can be further evaluated. After data collection, remove noise data, fill in missing values, and normalize the data.

[0053] Using SQL query to extract user historical purchase records from the e-commerce platform database can accurately obtain user purchase behavior data, identify user preferences and behavior patterns, improve the accuracy of recommendations, thereby enhancing user experience and conversion rate. Regularly obtaining user interaction data on social platforms through web crawling technology breaks the limitation of traditional e-commerce data being single, provides multi-dimensional user behavior data, and helps analyze user social influence and content preferences. By integrating the API of the live platform, real-time capture and analysis of user behavior during the live broadcast can respond to users' immediate needs, adjust push content and products in real-time, enhance the timeliness of product promotion, and perform data cleaning and preprocessing on the collected data to ensure data consistency and accuracy, avoiding affecting analysis results due to data quality problems.

[0054] Furthermore, extract the key features of users from the cleaned data, including users' interest features , purchase habit features and social influence ;

[0055] The extraction of interest degree features is to perform aggregation analysis on the historical purchase records of all users and calculate the interest degree of each user in different product categories ;

[0056] The calculation formula is:

[0057]

[0058] Among them, is the degree of interest of user u in product i, is the purchase frequency of user u for product i, is the total purchase frequency of user u;

[0059] For each user, an interest vector is constructed according to their degree of interest in different product categories , representing the distribution of the degree of interest of each user in each product category;

[0060] Interest feature is expressed as:

[0061]

[0062] Among them, is the degree of interest of user u in product i, reflecting the purchase frequency of the user in this category of products, is the total number of product categories;

[0063] The purchase habit feature , including the purchase frequency, purchase time period and purchase product price range features of the user;

[0064] The social influence is calculated by analyzing the interaction data (likes, comments, shares) of the user on the social platform and using the weighted summation method ,

[0065] The social influence calculation formula is:

[0066]

[0067] Among them, , , respectively represent the number of likes, comments and shares of user u on the social platform, , , are the weight coefficients of each interaction behavior, which are set through data analysis;

[0068] Extract the keywords of the user's social interaction data through the TF-IDF method to help accurately classify the user and evaluate the weight of each keyword in the user interaction text. The TF-IDF weight calculation formula is:

[0069]

[0070] Among them, is the weight of the vocabulary in the document d, For vocabulary The word frequency in document d, calculated as the vocabulary The ratio of the number of times it appears in the document to the total number of words in the document, is the inverse document frequency;

[0071] The calculation formula is:

[0072]

[0073] where N is the total number of documents, is the number of documents containing the vocabulary , and IDF is used to measure the importance of the vocabulary in all documents. Frequently occurring vocabulary will have a lower IDF value;

[0074] Extract keywords for each user in the social platform and aggregate them to obtain the keyword vector of the user , which represents the attention distribution of the user on all products and topics;

[0075] Keyword vector The expression is:

[0076]

[0077] where, is the attention of user u on the keyword , reflecting the intensity of interest in the keyword, is the total number of keywords;

[0078] Normalize the interest feature vector , purchase habit characteristics , social influence and keyword vector and merge them using weighted summation. Divide users into product promotion groups T according to similar characteristics through the K-means clustering algorithm.

[0079] By comprehensively extracting the key features of users, the problem of relying solely on a single data source is effectively avoided, making the user portrait more diverse and accurate, which helps to achieve more personalized recommendations, improve the efficiency and accuracy of recommendations. Through the quantitative calculation of the degree of interest, the invention can more accurately identify the interest preferences of users, avoid the inaccurate recommendations caused by rough classification, refine the purchase habits of users, and the invention can accurately grasp the consumption patterns of users, avoid the blindness of recommendations based on a single feature, and use the TF-IDF method to extract keywords in social interaction data, which can accurately capture the attention points of users on social platforms, enhance the accuracy of user classification, and standardize the extracted information and merge it using the weighted summation method. Finally, the user groups are accurately divided through the K-means clustering algorithm to achieve targeted product promotion and avoid the inefficiency caused by extensive recommendations.

[0080] S2. Based on the divided promotion groups, the influence maximization algorithm and collaborative filtering technology are combined to optimize the promotion groups, and personalized product promotion strategies are formulated;

[0081] Specifically, the combined features are converted into elements in the rating matrix to construct the rating matrix , with the behavior user u as the row and the product i as the column;

[0082] Collect the frequency of product i appearing in the ratings of each user and calculate the TF-IDF weight based on the product popularity;

[0083] The formula for calculating the product popularity is:

[0084]

[0085] Where is the total number of ratings of product i, T is the product promotion group, is the product in the frequency of the rating behavior of user u;

[0086] The formula for calculating the TF-IDF weight is:

[0087]

[0088] Where is the TF-IDF weight of product i, is the product in the frequency of the rating behavior of user u, is the number of ratings of user u, that is, the total number of user behaviors, T is the product promotion group, is the popularity of product i (the number of ratings of product i by all users), Let \(f_i\) be the frequency of product \(i\) in a certain user's rating. The higher the relative frequency, the greater the influence of product \(i\) on this user. Let \(idf_i\) be the inverse document frequency (IDF) of popularity, which reflects the rarity of product \(i\) in \(T\). The IDF value of popular products is lower, while that of niche products is higher.

[0089] After calculating the TF-IDF weights of each product, the rating of product \(i\) is corrected. Through this correction, the bias of popular products in the final recommendation is reduced, enabling the recommendation system to consider niche products more.

[0090] The formula for rating correction is as follows:

[0091]

[0092] Where, \(r_{ui}\) is the original rating of user \(u\) for product \(i\). \(w_i\) is the TF-IDF weight of product \(i\). The corrected rating matrix can more truly reflect the "preference intensity" of users for products rather than popularity.

[0093] Collect user characteristics, including the age, occupation, and gender of users, and calculate the similarity of user characteristics based on fuzzy logic to solve the similarity problem in the case of user cold start or incomplete user portraits, especially applicable in the initial stage of recommendation or when data is sparse, improving the accuracy of user similarity matching.

[0094] The formula for age similarity is as follows:

[0095]

[0096] Where, \(sim_{age}(u, r)\) is the age similarity between user \(u\) and user \(r\). \(|age_u - age_r|\) is the age difference between user \(u\) and user \(r\). When the age difference does not exceed 5 years, it is considered that the users are highly similar, and the similarity is 1. \(sim_{age}(u, r)\) is calculated as a gradually decreasing similarity when the age difference is between 5 and 25 years. The greater the age difference, the lower the similarity. When the age difference is greater than 25 years, the similarity is 0, that is, these two users are not considered similar.

[0097] The formula for occupation similarity is as follows:

[0098]

[0099] Where, \(sim_{occ}(u, r)\) is the occupation similarity between user \(u\) and user \(r\). \(d_{occ}(u, r)\) is the distance between user \(u\) and user \(r\) in the occupation tree. The occupation tree is a hierarchical structure, and the closer the occupations, the more similar they are. As a regulatory factor, it is used to control the influence degree of the occupational tree distance on the similarity. This value usually ranges from [0, 1]. A larger value indicates that more importance is attached to the occupational tree distance, and a smaller value indicates that users with smaller occupational differences can also be considered similar;

[0100] The gender similarity is calculated as follows:

[0101]

[0102] where is the gender similarity between user u and user r, , is the gender of user u, is the gender of user r. If the genders of user u and user r are the same, the gender similarity is 1; otherwise, the gender similarity is 0;

[0103] The age, occupation, and gender similarities are integrated into a comprehensive similarity using the weighted average method ;

[0104] An cognitive behavior propagation graph is constructed based on the propagation relationship between users and products. The nodes represent users and products, and the edges represent the behaviors of users towards products (such as rating, clicking, purchasing, etc.) or influence propagation;

[0105] Propagation influence The calculation formula is:

[0106]

[0107] where is the social influence, is the behavior response coefficient, which is set through experiments and represents the response degree of user r to the behavior of user u, is user The behavior intensity shown after user u recommends a product to user r. The larger this value, the stronger the user's cognition or interest in the product;

[0108] Based on the propagation influence, the influence maximization algorithm based on the greedy algorithm is used to find the user who has the greatest influence on the behaviors of other users. The greedy algorithm selects the user who can maximize the propagation influence at each step, thereby gradually expanding the propagation range. Through the propagation influence Select the user with the greatest propagation influence as the initial seed user set S. In each round of iteration, select the user u who can maximize the influence propagation currently, so that the influence increment in the entire propagation graph is the largest after adding this user;

[0109] Based on the propagation influence, calculate the influence increment of each user:

[0110]

[0111] Among them, is the propagation influence of the initial seed user set S, is the propagation influence after adding user u to the initial seed user set S, is the influence increment brought by user u;

[0112] Users are sorted in descending order according to the influence increment, and the user with the largest influence increment in the social network is selected as the propagation node for propagation. In each round of optimization, users are added in descending order to increase the influence in the propagation network until the user influence increment is less than the preset threshold (controlling the minimum growth of the propagation increment), then the contribution of the propagation influence to the final effect is the smallest, and the iterative optimization process stops;

[0113] By calculating the TF-IDF score similarity and the propagation influence similarity, and performing a weighted sum with the user feature similarity a comprehensive user similarity is obtained

[0114] The TF-IDF similarity calculation formula is:

[0115]

[0116] Among them, is the similarity between user u and user r, is the score of product i given by user u, , is the TF-IDF weight of product i, reflecting the scarcity and popularity of product i;

[0117] The propagation influence similarity calculation formula is:

[0118]

[0119] Among them, is the propagation influence of user u, is the propagation influence of user r, is the maximum value of all users' influence, reflecting the degree of closeness of the propagation influence difference. The more similar the propagation influence of users, the more similar the recommended products;

[0120] Use the K-Means++ algorithm to cluster the comprehensive user similarity and reassign a group to each user , remove the users in the assigned group from all users, and perform clustering again based on the remaining user group until all users are completely clustered and classified, and the users within the group have consistent similarity.

[0121] Set the comprehensive user similarity The threshold is b. When there is a comprehensive user similarity greater than or equal to b within the group, then the users in the group are the core user group. If the comprehensive user similarity within the group is less than b, then the users in the group are the potential user group.

[0122] By calculating the product popularity and TF-IDF weights, the scarcity of products and user preferences are quantified, so as to adjust the scoring matrix to optimize recommendations. The corrected scoring matrix can better reflect users' preferences for unpopular products, avoid over-recommendation of popular products, and collect multi-dimensional features of users (age, occupation, gender), and use fuzzy logic weighted calculation of similarity to solve the cold start problem faced by the recommendation system, ensuring that even in the case of sparse data, recommendations can still be effectively made based on users' basic characteristics, avoiding the inaccurate recommendations caused by insufficient information in traditional methods. By constructing a propagation graph and calculating the propagation influence, the influence of key users in the social network is identified and optimized, and the greedy algorithm is used to select the users with the greatest influence to achieve a wider and more effective product propagation, enhancing the influence of social recommendations, helping to quickly improve users' awareness of new products, enhancing the social effect of recommendations, and refining user classification through weighted calculation of comprehensive similarity to ensure that each user gets more accurate recommendations based on multiple dimensions (product preference, propagation influence, feature similarity). The clustering process is optimized through the K-Means++ algorithm, improving the accuracy of group division, making recommendations more personalized and efficient, and performing group screening to effectively improve the accuracy of recommendations, not only providing accurate content for core users, but also attracting potential users, enhancing the coverage and effect of the entire recommendation system.

[0123] Furthermore, promotion strategies are formulated separately for the core user group and the potential user group. For the core user group, membership pricing and differential pricing are set, and the influence of the core user group is used for social sharing rewards and recommendation reward mechanisms to encourage core users to promote products, and bundled sales and package discounts are launched to increase the purchase volume of core users. For the potential user group, by providing time-limited offers and group purchase discounts, users are encouraged to purchase and share products, and rewards can be obtained for successful promotion and purchase, and the point and cashback system is used to motivate users to purchase and participate in social sharing.

[0124] By offering special price concessions and customized pricing plans to the core user group, it is possible to effectively increase the purchase frequency and brand loyalty of core users. At the same time, it encourages them to engage in more social dissemination, leading to more user conversions, helping to expand the product's market share, increase sales volume and user stickiness. Through the social sharing and referral reward mechanism, it effectively stimulates the enthusiasm of core users, forms a self-spreading effect, improves the market penetration rate of the product, reduces the cost of traditional advertising promotion, improves the efficiency and accuracy of marketing. Combining bundling sales and package discounts can effectively increase product sales volume, and also enhance the overall purchase experience of users, making them feel more benefits and value when purchasing, and further enhancing user loyalty and long-term value. For the time-limited offers and group-buying discounts for potential user groups, it can effectively attract potential users to convert into paying users and encourage them to share the product with more people, thereby expanding the product's market share and enhancing the brand's popularity. Through the points and cashback system, it can stimulate users' repeat purchase behavior, not only increasing user activity and participation, but also enhancing the brand's customer stickiness, and ultimately helping to establish a stable user group and improve the long-term benefits of marketing.

[0125] S3. When implementing product promotion, by collecting and analyzing real-time user behavior data, using the particle swarm optimization algorithm combined with multivariate regression analysis and time series analysis to predict future market demand and cost changes, and adjusting product pricing and promotion strategies in real time.

[0126] Specifically, label the core user group as group A and the potential user group as group B. Respectively collect the purchase data of users when the strategies of groups A and B are implemented, and calculate the purchase conversion rates of users in groups A and B;

[0127] The purchase conversion rate of the users in group A Calculation formula:

[0128]

[0129] Wherein, is the purchase conversion rate of group A, is the number of users who make purchases in group A, and A is the core user group;

[0130] The purchase conversion rate of the users in group B Calculation formula:

[0131]

[0132] Wherein, is the purchase conversion rate of group B, is the number of users who make purchases in group B, and B is the potential user group;

[0133] When Then the conversion rate of Group B is higher than that of Group A. It is necessary to adjust the promotion strategy of Group A, increase the discount rate of Group A and launch time-limited preferential activities. When this happens, adjust the promotion strategy of Group B, extend the preferential time and increase the discount rate, and encourage users in Group B to purchase through group buying.

[0134] By collecting the purchase data of the core user group (Group A) and the potential user group (Group B), the behavioral responses of the two groups of users under different promotion strategies can be understood in detail, avoiding the blindness of a single strategy. By calculating the purchase conversion rates of Group A and Group B, the response degrees of the two groups of users under the current strategy can be quantitatively analyzed, and which ones need to be further optimized, avoiding subjective assumptions about the strategy effects, thereby improving the marketing efficiency and resource utilization rate. And by adjusting the promotion strategy of Group A, the invention can effectively improve the purchase conversion rate of core users, avoid too lenient promotion strategies for the core user group, and further maximize the purchase potential of existing users. By adjusting the promotion strategy of Group B, the invention can effectively improve the purchase conversion rate of potential users and narrow the conversion gap with core users.

[0135] Furthermore, use time series analysis to predict the market demand in the future for a period of time;

[0136] The time series analysis formula is:

[0137]

[0138] Among them, is the predicted market demand at time point z, is the constant term, representing the baseline sales volume, is the trend coefficient, set through genetic algorithm, representing the growth trend of sales volume, m is the time variable, is the error term, representing other unforeseen influencing factors;

[0139] Use multivariate regression analysis to predict the cost changes of products, including production costs, logistics costs, warehousing costs, etc., and adjust the pricing strategy to ensure maximum profit can still be obtained under the condition of cost changes;

[0140] The regression analysis calculation formula is:

[0141]

[0142] Among them, is the product cost at time point z, is the constant term of the regression, , , are the variables affecting the cost (such as production scale, raw material price, etc.), , , are the regression coefficients of each variable, which are set by simulated annealing and represent the influence degree of each factor on the cost;

[0143] The user behavior, future market demand, and cost changes are combined with the sparrow search algorithm through particle swarm optimization for multi-objective pricing optimization, which helps the particle swarm optimization algorithm maintain the diversity of optimization solutions during the multi-objective optimization process, avoid falling into local optima, and adjust the product pricing and promotion intensity in real time to improve the sales conversion rate and set the base price of each product , and this price is determined by market analysis and product cost. The adjustment factors are set as the popularity value (measuring the market popularity of the product based on data such as the product's view volume, click-through rate, and addition to the shopping cart) and user interest (measuring the user's interest in the product based on the user's purchase history, ratings, and other behaviors of the product);

[0144] The multi-objective pricing optimization formula is:

[0145]

[0146] Among them, is the price after dynamic adjustment, is the base price of the product, are the weight coefficients of popularity and interest, which are set by machine learning, is the popularity value of the product, is the user interest, is the weight coefficient of the price adjustment range, which is set by Bayesian optimization, is the actual amplitude of the price adjustment, indicating the change relative to the base price, is the weight coefficient of the market demand, which is set by fuzzy logic, is the weight coefficient of the product cost, which is set by an adaptive mechanism, is the weight coefficient of the profit target, which is set by deep learning methods, is the predicted market demand at time point z, is the product cost at time point z, is the profit target at time point z;

[0147] And according to the price after dynamic adjustment The particle swarm optimization algorithm is used to continuously optimize the product pricing. Each particle represents a pricing scheme, and the goal of the particle is to optimize the pricing strategy to improve the conversion rate. The particle continuously updates its speed and position , and find the optimal solution (the optimal product pricing and promotion strategy),

[0148] The calculation formulas for the update speed and position are as follows:

[0149]

[0150]

[0151] Among them, is the inertia weight, which is set through reinforcement learning, , is the acceleration constant, , is a random value, is the individual best position of the particle, is the global best position, is the position of the particle, and t is the number of iterations.

[0152] Through time series analysis, it provides a scientific prediction of future demand based on historical data, can identify trends and periodic changes in market demand, provides more accurate demand planning for product pricing. Through regression analysis, it accurately captures the driving factors of cost changes and adjusts the price strategy in real time in the case of cost fluctuations, ensuring that the product can maintain competitiveness in market changes, avoiding profit shrinkage caused by cost fluctuations, maximizing business profits, and combining the particle swarm optimization algorithm and the sparrow search algorithm for multi-objective optimization pricing, which can achieve more dynamic and accurate price adjustment, avoiding the limitations of fixed pricing strategies, enabling the product to flexibly respond to demand fluctuations in a complex market environment, improving the sales conversion rate and market competitiveness. Once again, using the particle swarm optimization algorithm, by simulating the search process of particles in nature, the pricing strategy is continuously iteratively updated to find the optimal pricing plan, ensuring that the pricing plan for each product always remains optimal in the changing market environment, which can effectively improve the sales conversion rate, while reducing the complexity and errors of manual pricing adjustment and enhancing the pricing efficiency.

[0153] This embodiment also provides a live e-commerce product promotion system, including the following steps:

[0154] A data collection module, which is used to obtain user data from e-commerce platforms, social media, and live API, and perform noise removal, missing value filling, and data normalization;

[0155] A group division module, which is used to extract keywords of users' interests, purchase habits, social influence, and social interaction data, and perform processing and merging, and conduct clustering to divide the product promotion groups;

[0156] A group optimization module, which is used to collect user characteristics using the product promotion groups, calculate the similarity between users, and optimize the promotion groups through greedy algorithms and collaborative filtering techniques;

[0157] A strategy formulation module for providing exclusive pricing and referral rewards for core users, and providing time-limited offers, group-buying discounts, and point rewards for socially active users;

[0158] A dynamic pricing module for predicting market demand and costs, adjusting pricing through a particle swarm optimization algorithm, and optimizing the sales conversion rate in real time.

[0159] This embodiment also provides a computer device applicable to the case of the live e-commerce product promotion method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the live e-commerce product promotion method proposed in the above embodiment.

[0160] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the outer shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0161] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for promoting live e-commerce products as 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 for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc.

[0162] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for promoting live e-commerce products, characterized in that: including Collecting key data of users through multiple channels and preprocessing it, and dividing the product promotion groups by using keyword search and social influence evaluation on the processed data; Based on the divided promotion groups, combining the influence maximization algorithm and collaborative filtering technology to optimize the promotion groups, and formulating personalized product promotion strategies; When implementing product promotion, by collecting and analyzing the real-time behavior data of users, using the particle swarm optimization algorithm combined with multivariate regression analysis and time series analysis to predict future market demand and cost changes, and adjusting the product pricing in real time.

2. The method for promoting live e-commerce products according to claim 1, wherein: The division of the product promotion group by using keyword search and social influence assessment for the processed data refers to extracting the key features of users from the cleaned data, including the interest features of users , purchase habit features and social influence ; Extract keywords of the user's social interaction data through the TF-IDF method and evaluate the weight of each keyword in the user interaction text ; Extract keywords for each user in the social platform and aggregate them to obtain the keyword vector of the user ; Interest features , purchase habit features , social influence and keyword vectors are standardized and combined using weighted summation, and the user product promotion groups T are divided according to similar features through the K-means clustering algorithm.

3. The method for promoting live e-commerce products according to claim 2, characterized in that: The optimization of the promotion groups by combining the influence maximization algorithm and collaborative filtering technology based on the divided promotion groups includes: Convert the combined features into elements in the scoring matrix to construct the scoring matrix where the rows represent users u and the columns represent products i; Collect the frequency of product i in the ratings of each user and calculate the TF-IDF weight based on product popularity ; After calculating the TF-IDF weights of each product, the score of product i is corrected to obtain the corrected score ; Collecting user characteristics, including the age, occupation and gender of users, and calculating the similarity of user characteristics based on fuzzy logic; Use the weighted average method to fuse age, occupation, and gender similarity into a comprehensive similarity ; Construct a cognitive behavior propagation graph based on the propagation relationship between users and products. Nodes represent users and products, and edges represent users' behaviors towards products, and calculate the propagation influence ; Based on the propagation influence, use the influence maximization algorithm based on the greedy algorithm to find the users with the greatest influence on the behavior of other users, and through the propagation influence Select the users with the greatest propagation influence as the initial seed user set S. In each round of iteration, select the user u that can maximize the influence propagation currently; Based on the propagation influence, calculate the influence increment of each user, sort the users in descending order according to the influence increment, select the user with the largest influence increment in the social network as the propagation node for propagation, and in each round of optimization, add users in descending order to increase the influence in the propagation network until the influence increment of the user is less than the preset threshold , then the contribution of the propagation influence to the final effect is the smallest, and the iterative optimization process stops; By calculating the TF-IDF score similarity and the propagation influence similarity, and performing weighted summation with the user feature similarity, a comprehensive user similarity is obtained , Use the K-Means++ algorithm for the comprehensive user similarity to perform clustering and reassign a group to each user , remove the users in the assigned groups from all users, and perform clustering again based on the remaining user groups until all users have completed clustering and classification; Set the comprehensive user similarity The threshold is b. When in the group there is a comprehensive user similarity greater than or equal to b, then the users in the group are the core user group. If the comprehensive user similarity within the group is less than b, then the users in the group are the potential user group.

4. The method for promoting live e-commerce products according to claim 3, wherein: The formulation of personalized product promotion strategies means formulating promotion strategies for the core user group and potential user group respectively. For the core user group, setting membership pricing and differential pricing, using the influence of the core user group for social sharing rewards and referral reward mechanisms to encourage core users to promote products, and launching bundled sales and package discounts. For the potential user group, by providing time-limited offers and group purchase discounts, encouraging users to purchase and share products, and getting rewards for successful promotion and purchase, and using the integral and cashback system to motivate users to purchase and participate in social sharing.

5. The method for promoting live e-commerce products according to claim 4, wherein: When implementing product promotion, collecting and analyzing real-time user behavior data for promotional strategy adjustment means labeling the core user group as Group A and the potential user group as Group B, respectively collecting the purchase data of users when the strategies of both Group A and Group B are implemented, and calculating the purchase conversion rate of Group A , calculating the purchase conversion rate of Group B ; When the conversion rate of Group B is higher than that of Group A, it is necessary to adjust the promotion strategy of Group A, increase the discount rate of Group A and launch limited-time preferential activities. When this happens, adjust the promotion strategy of Group B, extend the preferential time and increase the discount rate, and encourage users in Group B to purchase through group buying.

6. The method for promoting live e-commerce products according to claim 5, wherein: Using the particle swarm optimization algorithm combined with multivariate regression analysis and time series analysis to predict future market demand and cost changes, and adjusting product pricing in real time means using time series analysis to predict market demand in the next period of time ; Use multivariate regression analysis to predict the cost changes of products over a period of time in the future , including production costs, logistics costs, warehousing costs, etc., and adjust the pricing strategy; Multi-objective pricing optimization is carried out by combining the particle swarm optimization with the sparrow search algorithm for user behavior, future market demand, and cost changes , and the pricing and promotion intensity of products are adjusted in real time; And according to the dynamically adjusted price The particle swarm optimization algorithm is used to continuously optimize the product pricing. Each particle represents a pricing plan. The goal of the particle is to optimize the pricing strategy, and the particle continuously updates its velocity according to the current market environment and user behavior and position .

7. The method for promoting live e-commerce products according to claim 6, wherein: The collection of key data of users through multiple channels and preprocessing it means using SQL queries to extract the corresponding user historical purchase records from the e-commerce platform database, regularly obtaining the interaction data of users on the social platform through web crawling technology from the user's social media, using the API of the integrated live broadcast platform to regularly obtain the behavior data of users during the live broadcast. After data collection, removing noise data, filling missing values and normalizing the data.

8. A live e-commerce product promotion system, based on the live e-commerce product promotion method according to any one of claims 1 to 7, characterized in that: including A data collection module for obtaining user data from e-commerce platforms, social media and live broadcast APIs, and performing noise removal, missing value filling and data normalization; A group division module for extracting keywords of users' interests, purchase habits, social influence and social interaction data, and processing and merging them to perform clustering to divide product promotion groups; A group optimization module for using the product promotion groups to collect user characteristics, calculate the similarity between users, and optimize the promotion groups through the greedy algorithm and collaborative filtering technology; A strategy formulation module for providing exclusive pricing and referral rewards for core users, and providing time-limited offers, group purchase discounts and integral rewards for socially active users; A dynamic pricing module for predicting market demand and cost, adjusting the pricing through the particle swarm optimization algorithm, and optimizing the sales conversion rate in real time.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the live e-commerce product promotion method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the live e-commerce product promotion method according to any one of claims 1 to 7.

Images