A data processing method and system for a lead generation platform in the home furnishing industry

Abstract

This invention relates to the field of data processing technology and discloses a data processing method and system for a home furnishing industry traffic generation platform. The method includes acquiring and processing user behavior data to obtain user preference clusters; extracting association vectors of product style and material features based on the user preference clusters, calculating similarity to determine a preliminary matching product set; extracting high-frequency style and material combinations, and filtering to obtain a recommended product list; weighted fusing of user behavior data to determine a recommendation weight set and sorting to obtain a product recommendation sequence; if the user preference cluster update difference exceeds a threshold, updating the weights and adjusting the sequence; generating highly attractive summary text based on the optimized sequence, determining personalized traffic generation content, and synchronizing with the platform display logic. This method can improve recommendation accuracy, dynamic adaptability, and traffic generation effect.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a data processing method and system for a lead generation platform in the home furnishing industry. Background Technology

[0002] Data processing and user acquisition platforms play a crucial role in modern commerce within the home furnishing industry. By accurately analyzing user behavior and product characteristics, they drive personalized recommendations, significantly enhancing user experience and market competitiveness. With the deepening of digital transformation, the home furnishing industry's demand for data-driven marketing is becoming increasingly urgent. Platforms need to extract valuable information from massive amounts of data to achieve precise product recommendations and user acquisition.

[0003] In existing technologies, many platforms rely on traditional statistical analysis or simple rule matching for product recommendations. For example, they recommend similar products based solely on a user's browsing history, ignoring hidden style or material preferences within user tastes. These methods struggle to cope with the multidimensionality of user behavior and the complexity of product attributes. When processing high-dimensional, heterogeneous data, they often suffer from incomplete feature extraction, leading to style or material identification biases. Furthermore, they struggle to quickly adjust feature weights in dynamic data streams to adapt to changes in user preferences.

[0004] Therefore, existing technologies cannot meet the needs of home furnishing industry traffic acquisition platforms for precise and personalized recommendations, resulting in a low degree of matching between recommendation results and users' actual needs. Summary of the Invention

[0005] This invention provides a data processing method and system for a home furnishing industry traffic generation platform to solve the problem of low matching degree between recommendation results and users' actual needs.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a data processing method for a home furnishing industry lead generation platform, comprising: acquiring and processing user behavior data to obtain a user preference cluster; based on the user preference cluster, extracting association vectors of product style and material characteristics, calculating vector similarity, and determining a preliminary matching product set; extracting the most frequently occurring style and material combinations from the preliminary matching product set, filtering products from a pre-established product tag database and evaluation feedback dataset, and obtaining a recommended product list; performing weighted fusion on the user behavior data to determine a recommendation weight set and sorting the recommended product list to obtain a product recommendation sequence; if the update difference of the user preference cluster exceeds a preset update threshold, updating the recommendation weights, adjusting the product recommendation sequence, and obtaining an optimized product recommendation sequence; based on the optimized product recommendation sequence, parsing the style and material details of the products, generating highly attractive summary text, matching user preferences, and determining the final personalized lead generation content; and synchronizing the platform display logic in real time according to the final personalized lead generation content to obtain an updated configuration for the user interface.

[0007] In one optional implementation, the step of acquiring and processing user behavior data to obtain user preference clusters includes: acquiring real-time browsing history, favorites history, and shopping cart product information of users; collecting interaction frequency and product category data; filtering the latest behavior data to obtain a raw behavior dataset; performing data cleaning on the raw behavior dataset to remove missing values, outliers, and duplicate records to obtain the user behavior data; and grouping the user behavior data using the K-means clustering algorithm and determining user preference clusters by calculating the distance between user feature vectors.

[0008] In one optional implementation, the step of extracting association vectors of product style and material features based on the user preference cluster, calculating vector similarity, and determining a preliminary matching product set includes: obtaining product style and material feature data from the user preference cluster, performing data cleaning to remove noise data, and obtaining a cleaned preference dataset; generating embedding vectors of product style and material features through feature extraction based on the cleaned preference dataset, and obtaining a vector mapping set; if the dimension of the embedding vectors in the vector mapping set meets a preset dimension threshold, comparing the vectors in the vector mapping set by calculating cosine similarity to obtain a similarity score matrix; and obtaining the products corresponding to vectors with scores higher than a preset similarity threshold based on the similarity score matrix, and determining a preliminary matching product set.

[0009] In one optional implementation, the step of extracting the most frequent style and material combinations from the initially matched product set and filtering products from a pre-established product tag database and evaluation feedback dataset to obtain a recommended product list includes: performing frequency statistics on product features in the initially matched product set, extracting frequently occurring style preferences and material characteristic combinations to obtain a refined feature set; if the matching degree of style preferences and material characteristics in the refined feature set exceeds a preset matching degree threshold, then filtering products that match the style preferences and material characteristics from the pre-established product tag database and evaluation feedback dataset to obtain a candidate product set; based on the candidate product set, combined with browsing time data and click frequency records in the user behavior data, adjusting the recommendation priority using weighted sorting to determine the recommended product list.

[0010] In one optional implementation, the step of weighted fusion of the user behavior data to determine a recommendation weight set and sorting the recommended product list to obtain a product recommendation sequence includes: extracting recent browsing time, click frequency, favorites records, and shopping cart item information from the user behavior data; calculating the user behavior data using weighted fusion based on preset weight coefficients for each behavior indicator to obtain user preference scores for different products, thus forming a recommendation weight set; and sorting the recommended product list according to the recommendation weight set to obtain the product recommendation sequence.

[0011] In one optional implementation, the step of updating the recommendation weights and adjusting the product recommendation sequence to obtain an optimized product recommendation sequence if the user preference cluster update difference exceeds a preset update threshold includes: real-time monitoring of user preference cluster update data and calculating the feature difference value between the updated user preference cluster and the historical user preference cluster; if the feature difference value exceeds the preset update threshold, recalculating the preference score and generating new recommendation weights; and reordering the product recommendation sequence based on the new recommendation weights to form an optimized product recommendation sequence.

[0012] In one optional implementation, the step of analyzing the style and material details of products based on the optimized product recommendation sequence, generating highly attractive summary text, matching user preferences, and determining the final personalized traffic-driving content includes: for products in the optimized product recommendation sequence, analyzing the product description text, extracting the style and material details of the products, and obtaining trend keywords and tactile selling points; generating highly attractive summary text based on the trend keywords and tactile selling points, combined with preset scene association templates; if the matching degree between the highly attractive summary text and the user preference cluster is lower than a preset matching threshold, adjusting the scene association content and trend words, regenerating the summary text, until the matching degree reaches the standard, and determining it as the final personalized traffic-driving content.

[0013] Secondly, this invention provides a data processing system for a home furnishing industry lead generation platform, comprising: a data acquisition module for acquiring and processing user behavior data to obtain a user preference cluster; a preliminary matching module for extracting association vectors of product style and material characteristics based on the user preference cluster, calculating vector similarity, and determining a preliminary matched product set; a product filtering module for extracting the most frequently occurring style and material combinations from the preliminary matched product set, filtering products from a pre-established product tag database and evaluation feedback dataset, and obtaining a recommended product list; a sequence generation module for weighted fusion of the user behavior data, determining a recommendation weight set, and sorting the recommended product list to obtain a product recommendation sequence; a sequence optimization module for updating the recommendation weights and adjusting the product recommendation sequence if the update difference of the user preference cluster exceeds a preset update threshold, thereby obtaining an optimized product recommendation sequence; a content output module for parsing the style and material details of the products based on the optimized product recommendation sequence, generating highly attractive summary text, matching user preferences, and determining the final personalized lead generation content; and an interface configuration module for real-time synchronization of the platform display logic according to the final personalized lead generation content, thereby obtaining an updated configuration of the user interface.

[0014] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the data processing method for a home furnishing industry lead generation platform as described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the data processing method for a home furnishing industry lead generation platform described in any one of the above-mentioned methods.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] (1) This invention obtains user behavior data and user preference clusters, extracts the association vectors of product style and material features based on the user preference clusters and calculates the similarity to determine the initial matching product set. It can accurately capture users' potential preferences for home furnishing product style and material, solves the problem of the recommendation being out of touch with user needs caused by relying only on simple rule matching in the prior art, and improves the accuracy of product recommendation.

[0018] (2) This invention determines the recommendation weight set by weighted fusion of user behavior data, sorts the recommended product list to obtain the product recommendation sequence, and updates the recommendation weight and adjusts the sequence when the difference in user preference cluster updates exceeds the threshold, thereby realizing real-time response to the dynamic needs of users, overcoming the defect in the prior art that it is difficult to quickly adjust feature weights to adapt to changes in user preferences, and enhancing the dynamic adaptability of recommendations.

[0019] (3) Based on the optimized product recommendation sequence, the present invention generates highly attractive summary text, determines the final personalized traffic-driving content and synchronizes the platform display logic, which can present the accurately recommended products in a way that fits the user's preferences, improve the user's attention and acceptance of the traffic-driving content, and effectively improve the user experience and conversion rate of the home furnishing industry traffic-driving platform. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the data processing method for a customer acquisition platform in the home furnishing industry provided in the first embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the data processing system structure for a customer acquisition platform in the home furnishing industry, provided in the second embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Reference Figure 1 The first embodiment of the present invention provides a data processing method for a traffic generation platform in the home furnishing industry, comprising the following steps:

[0024] S11, acquire and process user behavior data to obtain user preference clusters;

[0025] S12, Based on the user preference cluster, extract the association vectors of product style and material features, calculate the vector similarity, and determine the preliminary matching product set;

[0026] S13, extract the most frequent style and material combinations from the initially matched product set, filter products from the pre-established product tag database and evaluation feedback dataset, and obtain a recommended product list;

[0027] S14, perform weighted fusion on the user behavior data, determine the recommendation weight set, and sort the recommended product list to obtain the product recommendation sequence;

[0028] S15, if the user preference cluster update difference exceeds the preset update threshold, update the recommendation weight, adjust the product recommendation sequence, and obtain an optimized product recommendation sequence;

[0029] S16, Based on the optimized product recommendation sequence, analyze the style and material details of the products, generate highly attractive summary text, match user preferences, and determine the final personalized traffic-driving content;

[0030] S17. Based on the final personalized traffic-driving content, the platform display logic is synchronized in real time to obtain the updated configuration of the user interface.

[0031] In step S11, user behavior data needs to be acquired and processed to obtain user preference clusters, including: acquiring real-time browsing history, favorites history, and shopping cart product information; collecting interaction frequency and product category data; filtering the latest behavior data to obtain the original behavior dataset; performing data cleaning on the original behavior dataset to remove missing values, outliers, and duplicate records to obtain the user behavior data; and using the K-means clustering algorithm to group the user behavior data and determining the user preference clusters by calculating the distance between user feature vectors.

[0032] It's important to note that user behavior data refers to various interactive records generated by users on the platform. This data reflects users' interests and needs. Real-time browsing records include information such as the URLs of product pages viewed and browsing timestamps; shopping cart information covers data on items added to the cart, such as the product name, model, price, and time of addition; interaction frequency refers to the number and frequency of times users click, favorite, or add items to their cart; and product category data is information on how products are categorized according to certain standards, such as sofas, beds, and dining tables. User feature vectors are vectors formed by quantifying this user behavior data. Each dimension represents a quantified value of a behavioral characteristic. By calculating the distance between different user feature vectors, the similarity of user preferences can be determined.

[0033] In this step, the first step is to retrieve users' real-time browsing records and purchase history from the database of the home furnishing industry traffic acquisition platform, while also collecting user interaction frequency and product category data. To ensure the timeliness of the data, behavioral data from the most recent period (e.g., the last 7 days) is filtered to form the raw behavioral dataset. Then, the raw behavioral dataset undergoes data cleaning, a crucial step in ensuring data quality. Specifically, the data is checked for missing values ​​(e.g., a record with an empty browsing time); outliers (e.g., negative purchase amounts or browsing durations exceeding a reasonable range, such as more than 24 hours); and duplicate records (i.e., the same user's behavior towards the same product being recorded multiple times). Data that does not meet these requirements is deleted, resulting in clean and accurate user behavior data.

[0034] In one implementation, the K-means clustering algorithm is used to group user behavior data. K-means clustering is a commonly used unsupervised learning algorithm that divides samples in a dataset into K distinct clusters, resulting in high similarity among samples within a cluster and low similarity among samples between clusters. In this step, after converting the user behavior data into user feature vectors, the Euclidean distance between these vectors is calculated to measure the similarity between users, thereby dividing users into different user preference clusters. Each cluster represents a group of users with similar preferences.

[0035] It's worth noting that filtering for the latest behavioral data ensures that the analyzed data reflects users' recent preferences, avoiding misjudgments of current user needs due to outdated data. Data cleaning removes noisy data, improving the accuracy of subsequent data analysis and processing. The clustering effect of the K-means clustering algorithm is influenced by the choice of K value. In practical applications, a suitable K value needs to be determined through multiple experiments based on the platform's user base and data characteristics to ensure that the partitioning of user preference clusters is reasonable and effective.

[0036] For example, suppose on a home furnishing platform, user A browsed the fabric sofa page three times in the past 7 days, with each browsing session lasting 2 minutes, 3 minutes, and 5 minutes respectively, and purchased a modern minimalist fabric sofa on August 12th; user B browsed the solid wood dining table page twice in the past 7 days, added one Nordic-style dining table to their favorites, but did not add it to their shopping cart. After collecting this data, the latest data is filtered to form the raw behavioral dataset. During the data cleaning process, it was found that a browsing timestamp was missing in one of user A's browsing records, so it was deleted; all of user B's data was complete and reasonable, so it was retained. Then, the behavioral data of users A and B are transformed into feature vectors. Let's assume user A's feature vector is [fabric sofa viewed 3 times, added to cart 1 time, click frequency 0.5 times / hour], and user B's feature vector is [solid wood dining table viewed 2 times, added to favorites 1 time, click frequency 0.3 times / hour]. Using the K-means clustering algorithm with K=2, the Euclidean distance is calculated to group user A and other users with similar sofa preferences into one cluster, and user B and other users with similar dining table preferences into another cluster, thus obtaining two different user preference clusters.

[0037] In step S12, it is necessary to extract the association vectors of product style and material features based on the user preference cluster, calculate the vector similarity, and determine the preliminary matching product set. This includes: obtaining product style and material feature data from the user preference cluster, cleaning the data to remove noise, and obtaining a cleaned preference dataset; generating embedding vectors of product style and material features through feature extraction based on the cleaned preference dataset, and obtaining a vector mapping set; if the dimension of the embedding vectors in the vector mapping set meets a preset dimension threshold, comparing the vectors in the vector mapping set by calculating cosine similarity to obtain a similarity score matrix; and obtaining the products corresponding to vectors with scores higher than a preset similarity threshold based on the similarity score matrix, thus determining the preliminary matching product set.

[0038] It's important to note that product style characteristics refer to the design style attributes of home furnishing products, such as Nordic style, modern minimalist style, and traditional Chinese style; material characteristics refer to the material attributes used in the product, such as solid wood, metal, glass, and fabric. These characteristics are important markers for distinguishing different home furnishing products and are also important factors for users when choosing products. Vector embedding is a method of transforming high-dimensional sparse feature data into low-dimensional dense vectors. It can capture the semantic relationships between features, making similar features closer together in the vector space. Cosine similarity is an index that measures the directional similarity between two vectors. Its value ranges from [-1, 1], and the closer the value is to 1, the more similar the directions of the two vectors are.

[0039] In this step, the style and material characteristics of the products that users are interested in are first extracted from the user preference cluster obtained in step S11. This data may come from product label information, descriptive text, etc. Then, this data is cleaned to remove noisy data, such as incorrect style labels (mislabeling modern minimalist style as Nordic style) and vague material descriptions (such as "composite material" without specifying the specific components), etc., to obtain a cleaned preference dataset.

[0040] In one implementation, the cleaned preference dataset needs to be processed using feature extraction (such as natural language processing models like Word2Vec and GloVe) to convert product style and material features into embedding vectors, with each feature corresponding to one embedding vector, thus forming a vector mapping set. Next, the dimensionality of the embedding vectors in the vector mapping set is checked to see if it meets a preset dimensionality threshold (e.g., 3D). If it does, the cosine similarity between the vectors is calculated. By calculating the cosine similarity between all vectors, a similarity score matrix is ​​constructed, where each element represents the similarity score between two corresponding vectors. Finally, based on the similarity score matrix, products corresponding to vectors with scores higher than a preset similarity threshold (e.g., 0.85) are selected. These products have a high similarity to the features in the user preference cluster, thus forming a preliminary matched product set.

[0041] It's worth noting that cleaning product style and material feature data ensures the accuracy of subsequent feature extraction and similarity calculation, preventing mismatch results due to erroneous data. The dimension setting of the embedding vector needs to be determined based on the complexity of the features and the amount of data; an appropriate dimension can better capture the correlation between features. The choice of a preset similarity threshold affects the size of the initial matching product set. A threshold that is too high may result in too few matched products; a threshold that is too low may introduce too many irrelevant products. Adjustments need to be made based on actual business needs.

[0042] For example, style and material feature data of products are extracted from the "Nordic style solid wood furniture preference cluster". This data may include features such as "Nordic style, solid wood, oak, white". During data cleaning, a data entry with a material labeled "unknown" was found and deleted; "Nordic style" was uniformly corrected to "Nordic style" to ensure data consistency. Then, natural language processing was used to transform these features into 3-dimensional embedding vectors. For example, Nordic style is mapped to vector [0.8, 0.2, 0.1], and solid wood material is mapped to [0.6, 0.4, 0.3], forming a vector mapping set. It was found that the dimensions of these embedding vectors are all 3-dimensional, which meets the preset dimension threshold. If a vector is only 2-dimensional due to missing data, it is removed to ensure calculation consistency. The cosine similarity between these vectors is calculated to obtain a similarity score matrix. For example, if the similarity score between "Nordic style" and a product feature vector labeled "Nordic style solid wood dining table" is 0.92, which is higher than the preset similarity threshold of 0.85, then this dining table is included in the initially matched product set.

[0043] In step S13, it is necessary to extract the most frequent style and material combinations from the initially matched product set, and filter products from the pre-established product tag database and evaluation feedback dataset to obtain a recommended product list. This includes: performing frequency statistics on product features in the initially matched product set, extracting frequently occurring style preferences and material characteristic combinations to obtain a refined feature set; if the matching degree of style preferences and material characteristics in the refined feature set exceeds a preset matching degree threshold, then filtering products that match the style preferences and material characteristics from the pre-established product tag database and evaluation feedback dataset to obtain a candidate product set; based on the candidate product set, combined with browsing time data and click frequency records in the user behavior data, using weighted sorting to adjust the recommendation priority, and determining the recommended product list.

[0044] It should be noted that the frequency statistics of product features refer to counting the number of times various style and material combinations appear in the initially matched product set. By statistically analyzing these combinations, we can identify the most popular style and material pairings within the user preference cluster. The matching degree between style preferences and material characteristics is an indicator of the rationality and harmony between the two. For example, Nordic style is usually well-matched with solid wood and fabrics, resulting in a high matching degree; while the matching degree between Chinese classical style and metal materials is relatively low.

[0045] The pre-established product label database is a database that stores various label information for products, including labels such as style, material, color, and size; the evaluation feedback dataset contains information such as user evaluations, ratings, and usage feedback on the products, which can reflect the product quality and user satisfaction.

[0046] In this step, the style and material characteristics of each product in the initially matched product set are first extracted, and then the frequency of each style and material combination is counted. For example, if the combination of "Nordic style + solid wood" appears 20 times and "modern minimalist + fabric" appears 15 times, the most frequent combinations are extracted to form a refined feature set. Next, the matching degree between each style preference and material characteristic in the refined feature set is calculated, and the total number of occurrences of all material characteristic combinations under each style preference is counted. The proportion of the former to the latter is calculated as the matching degree. If the matching degree exceeds a preset matching degree threshold (e.g., 0.8), products with these style and material tags are selected from the product tag database. At the same time, referring to the evaluation feedback dataset, products with higher user ratings (e.g., ratings of not less than 4.5) and better reviews are selected to form a candidate product set.

[0047] In one implementation, browsing duration and click frequency data from user behavior data are also combined to perform a weighted sorting of the candidate product set. For example, the weight of browsing duration is set to 0.6 and the weight of click frequency is set to 0.4. A comprehensive score is calculated for each product, and the products are sorted from high to low based on the comprehensive score. The recommendation priority is adjusted, and the products with higher rankings are recommended first, thereby determining the recommended product list.

[0048] It's worth noting that extracting frequently occurring style and material combinations allows us to focus on the most mainstream needs within user preference clusters, improving the targeting of recommendations. Referencing review and feedback data when filtering products ensures that recommended products have good quality and user acceptance, enhancing the user experience. The weighting settings for weighted sorting need to be adjusted based on the platform's business objectives and user behavior characteristics.

[0049] For example, the initial matching product set contains 50 products. After frequency analysis of style and material combinations, the "Nordic style + oak" combination appears 17 times, making it the most frequent combination. This combination is then included in the refined feature set. The matching degree of this combination is calculated. For instance, if the total number of material combinations under "Nordic style" is 20, then the matching degree of "Nordic style + oak" is 17 / 20 = 0.85, exceeding the preset matching degree threshold of 0.8. All products labeled "Nordic style" and "oak" are selected from the product tag database, totaling 80 products. Then, products with user ratings of at least 4.5 are selected from the review feedback dataset, resulting in 60 products, forming the candidate product set. Combining user behavior data, a Nordic style oak dining table has a cumulative browsing time of 100 minutes and a click frequency of 5 times; another Nordic style oak dining chair has a cumulative browsing time of 80 minutes and a click frequency of 8 times. We set a weight of 0.6 for browsing time and 0.4 for click frequency, with a total browsing time of 500 minutes and a total click frequency of 20. We calculated the overall scores for the two products: the dining table's overall score was (100 / 500) × 0.6 + (5 / 20) × 0.4 = 0.22; the dining chair's overall score was (80 / 500) × 0.6 + (8 / 20) × 0.4 = 0.256. Based on the overall scores, the dining chair had a higher score. Therefore, we sorted the candidate products and adjusted the recommendation order of the dining chair to be placed in front of the dining table, resulting in the recommended product list.

[0050] In step S14, the user behavior data needs to be weighted and fused to determine the recommendation weight set and sort the recommended product list to obtain the product recommendation sequence. This includes: extracting the user's recent browsing time, click frequency, favorites records, and shopping cart product information from the user behavior data; calculating the user behavior data using weighted fusion based on the preset weight coefficients of each behavior indicator to obtain the user's preference scores for different products, thus forming the recommendation weight set; and sorting the recommended product list according to the recommendation weight set to obtain the product recommendation sequence.

[0051] It's important to note that recent browsing time refers to the total time a user spends browsing a product page within a recent period; longer browsing time generally indicates higher user interest in the product. Click frequency is the ratio of the number of times a user clicks on a product to the total number of clicks, reflecting the user's level of attention to the product. Favorites records indicate products that a user has marked as favorites, directly reflecting their interest. Shopping cart information shows the user's purchase history for that product or similar products, reflecting their actual purchasing preferences. The weighting coefficients for each behavioral indicator are pre-set based on the degree to which different behaviors influence user preferences. For example, purchasing behavior most directly reflects user preferences, so its weighting coefficient can be set higher; while browsing time has a relatively lower weighting coefficient.

[0052] In this step, user behavior data is first extracted to include recent (e.g., the last 15 days) browsing time, click frequency, favorites history (whether the item was saved), and shopping cart information (whether it was added to the cart). Then, based on pre-defined weighting coefficients for each behavior indicator (e.g., 0.4 for adding to cart, 0.2 for favorites, 0.2 for click frequency, and 0.2 for browsing time), a weighted calculation is performed on each product's behavior indicators to obtain the user's preference score for that product. For example, if a user has added to cart history (recorded as 1), favorites history (recorded as 1), click frequency 0.3, and browsing time 10 minutes (normalized to 0.2), then the preference score for that product is 1×0.4 + 1×0.2 + 0.3×0.2 + 0.2×0.2 = 0.7. The preference scores for all products constitute the recommendation weight set. Finally, the recommended product list is sorted from high to low according to the preference scores of each product in the recommendation weight set to obtain the product recommendation sequence. The higher the preference score, the higher the position of the product in the sequence.

[0053] It's worth noting that weighted fusion of user behavior data comprehensively considers the impact of various user behaviors on preferences, avoiding biases caused by single behavioral indicators, and making the resulting preference scores more accurately reflect users' true needs. The weighting coefficients need to be adjusted based on the platform's actual data and business experience to improve the accuracy of recommendations.

[0054] For example, a recommended product list contains 10 products. Recent behavioral metrics of users towards these 10 products are extracted from user behavior data. Product C's behavioral metrics are: added to cart history, favorited record, click frequency 0.5, and browsing time 20 minutes (normalized to 0.4). Product D's behavioral metrics are: no added to cart history, favorited record, click frequency 0.6, and browsing time 25 minutes (normalized to 0.5). Preset weighting coefficients are: purchase history 0.4, favorited record 0.2, click frequency 0.2, and browsing time 0.2. Therefore, the preference score for product C is 1×0.4 + 1×0.2 + 0.5×0.2 + 0.4×0.2 = 0.78; the preference score for product D is 0×0.4 + 1×0.2 + 0.6×0.2 + 0.5×0.2 = 0.42. Based on preference scores, product C will appear earlier in the product recommendation sequence than product D.

[0055] In step S15, if the user preference cluster update difference exceeds a preset update threshold, the recommendation weight is updated, the product recommendation sequence is adjusted, and an optimized product recommendation sequence is obtained. This includes: real-time monitoring of the user preference cluster update data, calculating the feature difference value between the updated user preference cluster and the historical user preference cluster; if the feature difference value exceeds the preset update threshold, recalculating the preference score and generating a new recommendation weight; and reordering the product recommendation sequence based on the new recommendation weight to form an optimized product recommendation sequence.

[0056] It's important to note that updated user preference cluster data refers to data that changes the original user preference cluster characteristics as users continuously generate new behaviors on the platform (such as new browsing, adding to cart, and favorites). The feature difference value is a quantitative indicator used to measure the degree of difference in features between the updated user preference cluster and the historical user preference cluster. It can be obtained by calculating the Euclidean distance or cosine distance between their feature vectors. The preset update threshold is a critical value for determining whether user preferences have changed significantly. When the feature difference value exceeds this threshold, it indicates a significant change in user preferences, requiring adjustments to the recommendation strategy.

[0057] In this step, the platform's data monitoring mechanism first tracks changes in user behavior data in real time. When a user engages in new interactions (such as browsing new product categories or adding products with different styles to their cart), the user preference cluster is updated. Then, the feature difference between the updated user preference cluster and the historical user preference cluster before the update is calculated. Specifically, this is done by comparing the feature vectors of the two clusters and calculating the distance between them. If this feature difference exceeds a preset update threshold (e.g., 0.3), it indicates a significant change in the user's preferences, requiring a recalculation of the user's preference scores for each product in the recommended product list. During the recalculation, the same behavioral indicators and weighting coefficients as in step S14 are used, but updated user behavior data is employed to generate new recommendation weights. Finally, the original product recommendation sequence is reordered based on the new recommendation weights, placing products that better match the user's current preferences at the top, forming an optimized product recommendation sequence.

[0058] It's worth noting that real-time monitoring of updated user preference clusters can promptly capture dynamic changes in user preferences, ensuring that recommended content keeps pace with evolving user needs. Setting update thresholds avoids frequent recommendations due to accidental user behavior; updates are only performed when preferences change significantly, guaranteeing the stability and effectiveness of recommendations. The process of recalculating preference scores and adjusting the recommendation sequence ensures that the recommendation results remain consistent with the user's latest preferences, improving the user experience.

[0059] For example, a user preference cluster originally focused on furniture in the "modern minimalist + fabric" style, with a historical feature vector of [0.6, 0.3, 0.1] (corresponding to the weights of modern minimalist style, fabric material, and other features, respectively). After a period of time, the user frequently browsed furniture in the "Nordic style + solid wood" style, adding multiple related behavioral data points. The updated user preference cluster feature vector became [0.2, 0.2, 0.6] (corresponding to the weights of modern minimalist style, fabric material, and Nordic style + solid wood material, respectively). Calculating the Euclidean distance between the two feature vectors yielded a feature difference value of 0.7, exceeding the preset update threshold of 0.3, thus requiring an update to the recommendation weights. Recalculating the user's preference scores for each product in the recommended product list revealed a significant increase in the preference score for the previously low-ranked "Nordic style solid wood dining table," while the preference score for the "modern minimalist fabric sofa" decreased. Based on the new recommendation weights, the product recommendation sequence was reordered, with the "Nordic style solid wood dining table" now ranked higher, forming an optimized product recommendation sequence.

[0060] In step S16, based on the optimized product recommendation sequence, it is necessary to analyze the style and material details of the products, generate highly attractive summary text, match user preferences, and determine the final personalized traffic-driving content. This includes: for the products in the optimized product recommendation sequence, analyzing the product description text, extracting the style and material details of the products, and obtaining trend keywords and tactile selling points; generating highly attractive summary text based on the trend keywords and tactile selling points, combined with a preset scene association template; if the matching degree between the highly attractive summary text and the user preference cluster is lower than a preset matching threshold, adjusting the scene association content and trend words, regenerating the summary text, until the matching degree reaches the standard, and determining it as the final personalized traffic-driving content.

[0061] It's important to note that the product description text is a detailed introduction to the product on the platform, including style descriptions (such as "simple and smooth lines, showcasing Nordic style"), material descriptions (such as "made of imported oak, with a hard texture"), and functional features. Style tone refers to the design style and overall atmosphere embodied by the product, such as the fresh and natural Nordic style or the simple and clean modern minimalist style. Material details are specific descriptions of the materials used in the product, such as the type of wood or the texture of the fabric. Trend keywords are currently popular style and design-related terms in the home furnishing industry, such as "minimalist style" or "wabi-sabi aesthetics." Tactile selling points refer to the tactile experience of the product's materials, such as "delicate and smooth" or "soft and comfortable."

[0062] The preset scenario association templates are pre-designed text templates that combine products with everyday life scenarios, such as "living room scenario" and "bedroom scenario," helping users better imagine the product's practical application. Matching degree is an indicator that measures the degree of fit between highly attractive summary text and user preference clusters, obtained by calculating the overlap between keywords in the text and user preference features.

[0063] In this step, the product description text is first extracted for each item in the optimized product recommendation sequence, and then parsed using natural language processing techniques (such as word segmentation and keyword extraction). From the parsing results, words that reflect the product's style (such as "Nordic style" and "minimalist design") and words related to material details (such as "imported walnut wood" and "skin-friendly cotton and linen fabric") are extracted, and then trend keywords (such as "naturalism" and "light luxury texture") and tactile selling points (such as "warm and delicate" and "breathable and comfortable") are refined.

[0064] In one implementation, based on the extracted trend keywords and tactile selling points, a suitable preset scenario association template (such as the "living room leisure scene" template for sofas) needs to be selected. The keywords and selling points are then integrated into the template to generate highly attractive summary text. For example, combining the trend keyword "natural wood style" for "Nordic style solid wood sofa," the tactile selling point "warm wood touch," and the "living room leisure scene" template, the summary text generated is: "This Nordic style solid wood sofa exudes a fresh and natural wood style, and the warm wood touch brings a comfortable experience, creating a lazy and comfortable leisure space for your living room."

[0065] In addition, the matching degree between the summary text and the user preference cluster needs to be calculated. The user preference cluster contains user preference information on style, material, scene, etc. If the matching degree is lower than the preset matching threshold (e.g., 0.6), the scene association content (e.g., changing "living room leisure scene" to "family gathering scene") and trendy words (e.g., changing "natural wood style" to "Nordic minimalist style") are adjusted, the summary text is regenerated, and the matching degree is calculated again until the matching degree reaches the preset threshold. The summary text at this point is the final personalized traffic-driving content.

[0066] It's worth noting that analyzing product description text and extracting trendy keywords and tactile selling points allows for a precise grasp of the product's core characteristics, laying the foundation for generating attractive lead-generating content. Combining this with pre-set scenario-based templates makes the summary text more vivid, allowing users to easily empathize and increasing their interest in the product. By repeatedly adjusting the summary text to reach a matching threshold, it's possible to ensure that the lead-generating content highly aligns with user preferences, enhancing the lead generation effect.

[0067] For example, one product in the optimized product recommendation sequence is a "Japanese Wabi-sabi style ceramic vase," whose product description text is: "Made of coarse pottery, handcrafted, with simple lines, showcasing Japanese Wabi-sabi aesthetics. The surface has a natural, rough texture, suitable for displaying dried flowers, adding a tranquil atmosphere to your home." Analyzing this text, the style tone is extracted as "Japanese Wabi-sabi style" and "simple lines," and the material details are "coarse pottery material," "handcrafted," and "rough texture," yielding the trend keywords "Wabi-sabi aesthetics" and "handmade art," and the tactile selling point "natural roughness." Combining this with the "Bedroom Decoration Scene" template, the summary text is generated as follows: "This Japanese Wabi-sabi style ceramic vase, handcrafted from coarse pottery, with a natural, rough texture, fully showcases Wabi-sabi aesthetics, adding a tranquil atmosphere to your bedroom and is an excellent choice for displaying dried flowers."

[0068] The matching score between the text and the user preference cluster was calculated. The user preference cluster showed that the user likes a "natural and simple" style, with a matching score of 0.5, which is lower than the preset matching threshold of 0.6. The scene association content was adjusted to "study room decoration scene", and the trendy words were changed to "natural and simple style". The summary text was regenerated as follows: "This Japanese wabi-sabi style ceramic vase, with its handmade rough pottery material, brings a natural and rough touch. The natural and simple style fully demonstrates the wabi-sabi aesthetic, adding a tranquil atmosphere to your study room. Placing dried flowers on it makes it even more elegant." The matching score was calculated again and was 0.7, reaching the preset threshold. This text was determined to be the final personalized traffic-driving content.

[0069] In step S17, the platform display logic needs to be synchronized in real time according to the final personalized traffic-driving content to obtain the updated configuration of the user interface.

[0070] It's important to note that personalized content preferences in user historical interaction data include records of user actions such as clicks, dwell time, and sharing of promotional content. These records reflect users' acceptance and preferences for different types of promotional content. Platform display logic refers to the rules and methods for displaying platform pages, including page layout (such as the arrangement of product cards), content display order (such as the position of promotional content), and visual elements (such as background color and font size). Real-time synchronization means immediately applying the adjusted display logic to the platform to ensure users see the updated interface promptly.

[0071] This step begins by analyzing historical user interaction data to understand user preferences for different personalized content. For example, users tend to click on promotional content at the top of the page and prefer a clean page layout. Based on the characteristics of the final personalized promotional content (such as content length and style), the direction for adjusting the platform's display logic is determined. This includes placing the promotional content in a more prominent position on the page and using visual elements that match the content's style. Then, according to this adjustment direction, the platform's display logic is specifically configured: in terms of page layout, the display area for personalized promotional content is expanded; in terms of content display order, the promotional content is placed at the beginning of the page; and in terms of visual elements, colors (such as a light-colored background for Nordic-style products) and fonts (such as a sans-serif font for minimalist-style products) that match the product's style and tone are selected.

[0072] In one implementation, the configured display logic needs to be synchronized in real time to the front-end system of the home furnishing industry traffic generation platform using technical means to generate an updated configuration file for the user interface. When the user refreshes the page or logs back into the platform, the system loads this updated configuration file, and the user interface updates accordingly, displaying a new interface containing the final personalized traffic generation content.

[0073] It's worth noting that adjusting the display logic based on users' personalized content preferences allows platform interface updates to better align with user habits and visual preferences, increasing user engagement with the promotional content. Adjusting the page layout, content display order, and visual elements highlights personalized promotional content, enhancing its appeal and recognizability. Real-time synchronization of display logic and user interface updates ensures users see the latest promotional content promptly, improving promotional effectiveness and user experience.

[0074] For example, the final personalized lead generation content is a highly attractive summary text about "Nordic style solid wood desks." User interaction history data shows that users prefer clicking on lead generation content with images at the top of the page and prefer a light blue background. Therefore, the platform's display logic is adjusted as follows: place this lead generation content at the top of the page, paired with a high-resolution image of the desk, with a light blue background. Based on this adjustment, the platform's display logic is configured as follows: in terms of page layout, the lead generation content area is set as a banner at the top of the page, filling the entire page width; in terms of content display order, the lead generation content is placed before all product recommendations; in terms of visual elements, the background color is set to light blue, and the font uses a sans-serif font. These settings are synchronized to the platform in real time, generating an updated configuration for the user interface. After refreshing the page, users see the light blue banner area at the top of the page displaying the lead generation content and image of "Nordic style solid wood desks," which aligns with their usage habits and preferences.

[0075] In summary, this invention discloses a data processing method for a home furnishing industry traffic generation platform, comprising: acquiring user behavior data and user preference clusters; based on the user preference clusters, extracting association vectors of product style and material characteristics, calculating vector similarity, and determining a preliminary matching product set; extracting the most frequently occurring style and material combinations from the preliminary matching product set, filtering products from a pre-established product tag database and evaluation feedback dataset, and obtaining a recommended product list; performing weighted fusion on the user behavior data, determining a recommendation weight set, and sorting the recommended product list to obtain a product recommendation sequence; if the update difference of the user preference cluster exceeds a preset update threshold, updating the recommendation weights, adjusting the product recommendation sequence, and obtaining an optimized product recommendation sequence; based on the optimized product recommendation sequence, parsing the style and material details of the products, generating highly attractive summary text, matching user preferences, and determining the final personalized traffic generation content; and according to the final personalized traffic generation content, synchronizing the platform display logic in real time to obtain the updated configuration of the user interface. This invention acquires user behavior data and user preference clusters, extracts association vectors of product style and material features based on these clusters, calculates similarity, and determines a preliminary matching product set. This accurately captures users' potential preferences for home furnishing product styles and materials, solving the problem of recommendations being out of touch with user needs caused by relying solely on simple rule matching in existing technologies, thus improving the accuracy of product recommendations. Furthermore, by weighting and fusing user behavior data to determine a recommendation weight set, the recommended product list is sorted to obtain a product recommendation sequence. When the difference in user preference cluster updates exceeds a threshold, the recommendation weights are updated and the sequence is adjusted, achieving real-time response to dynamic user needs. This overcomes the deficiency in existing technologies where feature weights are difficult to quickly adjust to adapt to changes in user preferences, enhancing the dynamic adaptability of recommendations. In addition, this invention generates highly attractive summary text based on the optimized product recommendation sequence, determines the final personalized lead generation content, and synchronizes it with the platform display logic. This presents accurately recommended products in a way that aligns with user preferences, increasing user attention and acceptance of the lead generation content, effectively improving the user experience and conversion rate of home furnishing industry lead generation platforms.

[0076] Reference Figure 2The second embodiment of the present invention provides a data processing system for a home furnishing industry traffic generation platform, comprising: a data acquisition module for acquiring and processing user behavior data to obtain a user preference cluster; a preliminary matching module for extracting association vectors of product style and material features based on the user preference cluster, calculating vector similarity, and determining a preliminary matching product set; a product filtering module for extracting the most frequently occurring style and material combinations in the preliminary matching product set, filtering products from a pre-established product tag database and evaluation feedback dataset, and obtaining a recommended product list; a sequence generation module for weighted fusion of the user behavior data, determining a recommendation weight set, and sorting the recommended product list to obtain a product recommendation sequence; a sequence optimization module for updating the recommendation weights and adjusting the product recommendation sequence if the update difference of the user preference cluster exceeds a preset update threshold, thereby obtaining an optimized product recommendation sequence; a content output module for parsing the style and material details of the products based on the optimized product recommendation sequence, generating highly attractive summary text, matching user preferences, and determining the final personalized traffic generation content; and an interface configuration module for real-time synchronization of the platform display logic according to the final personalized traffic generation content, thereby obtaining an updated configuration of the user interface.

[0077] It should be noted that the data processing system for a home furnishing industry traffic generation platform provided in this embodiment of the invention is used to execute all the process steps of the data processing method for a home furnishing industry traffic generation platform described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0078] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data processing program for a home furnishing industry lead generation platform. When the processor executes the computer program, it implements the steps in the various data processing method embodiments for home furnishing industry lead generation platforms described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the sequence generation module.

[0079] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0080] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0081] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines.

[0082] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0083] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0084] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0085] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A data processing method for a customer acquisition platform in the home furnishing industry, characterized in that, include: Acquire and process user behavior data to obtain user preference clusters; Based on the user preference cluster, the association vectors of product style and material features are extracted, the vector similarity is calculated, and the preliminary matching product set is determined. Extract the most frequent style and material combinations from the initially matched product set, filter products from the pre-established product tag database and evaluation feedback dataset, and obtain a recommended product list; The user behavior data is weighted and fused to determine the recommendation weight set, and the recommended product list is sorted to obtain the product recommendation sequence; If the difference in the update of the user preference cluster exceeds a preset update threshold, the recommendation weight is updated, the product recommendation sequence is adjusted, and an optimized product recommendation sequence is obtained. Based on the optimized product recommendation sequence, the style and material details of the products are analyzed to generate highly attractive summary text, match user preferences, and determine the final personalized traffic-driving content. Based on the final personalized traffic-driving content, the platform display logic is synchronized in real time to obtain the updated configuration of the user interface; The step of acquiring and processing user behavior data to obtain a user preference cluster includes: Obtain real-time user browsing history, favorites history, and shopping cart product information; collect interaction frequency and product category data; filter the latest behavioral data; and obtain the raw behavioral dataset. For the original behavior dataset, data cleaning is performed to remove missing values, outliers, and duplicate records to obtain the user behavior data; The user behavior data is grouped using the K-means clustering algorithm, and user preference clusters are determined by calculating the distance between user feature vectors. The step of extracting association vectors of product style and material features based on the user preference cluster, calculating vector similarity, and determining a preliminary matching product set includes: Product style and material feature data are obtained from the user preference cluster, and data cleaning is performed to remove noise data, resulting in a cleaned preference dataset; Based on the cleaned preference dataset, embedding vectors of product style and material features are generated through feature extraction, resulting in a vector mapping set; If the dimension of the embedded vector in the vector mapping set meets the preset dimension threshold, then the vectors in the vector mapping set are compared by calculating the cosine similarity to obtain a similarity score matrix; Based on the similarity score matrix, obtain the products corresponding to the vectors with scores higher than a preset similarity threshold, and determine the initial matching product set; The optimized product recommendation sequence analyzes the style and material details of the products, generates highly attractive summary text, matches user preferences, and determines the final personalized traffic-driving content, including: For the products in the optimized product recommendation sequence, the product description text is parsed to extract the style and material details of the products, thereby obtaining trend keywords and tactile selling points; Based on the trend keywords and tactile selling points, and combined with preset scene association templates, highly attractive summary text is generated; If the matching degree between the highly attractive summary text and the user preference cluster is lower than the preset matching threshold, the scene association content and trendy words are adjusted, and the summary text is regenerated until the matching degree reaches the standard, and then determined as the final personalized traffic-driving content.

2. The data processing method for a home furnishing industry traffic generation platform according to claim 1, characterized in that, The process involves extracting the most frequent style and material combinations from the initially matched product set, filtering products from a pre-established product tag database and evaluation feedback dataset, and obtaining a recommended product list, including: Frequency statistics are performed on the product features in the initially matched product set to extract high-frequency style preferences and material characteristic combinations, resulting in a refined feature set. If the matching degree between style preference and material characteristics in the refined feature set exceeds the preset matching degree threshold, then products that match the style preference and material characteristics are selected from the pre-established product label database and evaluation feedback dataset to obtain a candidate product set. Based on the candidate product set, combined with the browsing time data and click frequency records in the user behavior data, a weighted sorting method is used to adjust the recommendation priority and determine the recommended product list.

3. The data processing method for a home furnishing industry traffic generation platform according to claim 1, characterized in that, The step of weighting and fusing the user behavior data to determine the recommendation weight set and sorting the recommended product list to obtain the product recommendation sequence includes: Extract recent browsing time, click frequency, favorites records, and shopping cart item information from the user behavior data; The user behavior data is calculated by weighted fusion based on the preset weight coefficients of each behavior indicator to obtain the user's preference score for different products, which constitutes the recommendation weight set. The recommended product list is sorted according to the recommended weight set to obtain the product recommendation sequence.

4. The data processing method for a home furnishing industry traffic generation platform according to claim 1, characterized in that, If the difference in user preference cluster updates exceeds a preset update threshold, the recommendation weights are updated, the product recommendation sequence is adjusted, and an optimized product recommendation sequence is obtained, including: Real-time monitoring of updated user preference clusters, and calculation of feature differences between the updated user preference clusters and historical user preference clusters; If the feature difference value exceeds the preset update threshold, the preference score is recalculated and a new recommendation weight is generated; The product recommendation sequence is reordered based on the new recommendation weights to form an optimized product recommendation sequence.

5. A data processing system for a lead generation platform in the home furnishing industry, characterized in that, include: The data acquisition module acquires and processes user behavior data to obtain user preference clusters. The preliminary matching module extracts the association vectors of product style and material features based on the user preference cluster, calculates the vector similarity, and determines the preliminary matching product set. The product filtering module extracts the most frequent style and material combinations from the initially matched product set, filters products from a pre-established product tag database and evaluation feedback dataset, and obtains a recommended product list. The sequence generation module performs weighted fusion on the user behavior data, determines the recommendation weight set, and sorts the recommended product list to obtain a product recommendation sequence. The sequence optimization module updates the recommendation weights and adjusts the product recommendation sequence if the user preference cluster update difference exceeds a preset update threshold, thereby obtaining an optimized product recommendation sequence. The content output module, based on the optimized product recommendation sequence, analyzes the style and material details of the products, generates highly attractive summary text, matches user preferences, and determines the final personalized traffic-driving content. The interface configuration module synchronizes the platform display logic in real time based on the final personalized traffic-driving content to obtain the updated configuration of the user interface. The step of acquiring and processing user behavior data to obtain a user preference cluster includes: Obtain real-time user browsing history, favorites history, and shopping cart product information; collect interaction frequency and product category data; filter the latest behavioral data; and obtain the raw behavioral dataset. For the original behavior dataset, data cleaning is performed to remove missing values, outliers, and duplicate records to obtain the user behavior data; The user behavior data is grouped using the K-means clustering algorithm, and user preference clusters are determined by calculating the distance between user feature vectors. The step of extracting association vectors of product style and material features based on the user preference cluster, calculating vector similarity, and determining a preliminary matching product set includes: Product style and material feature data are obtained from the user preference cluster, and data cleaning is performed to remove noise data, resulting in a cleaned preference dataset; Based on the cleaned preference dataset, embedding vectors of product style and material features are generated through feature extraction, resulting in a vector mapping set; If the dimension of the embedded vector in the vector mapping set meets the preset dimension threshold, then the vectors in the vector mapping set are compared by calculating the cosine similarity to obtain a similarity score matrix; Based on the similarity score matrix, obtain the products corresponding to the vectors with scores higher than a preset similarity threshold, and determine the initial matching product set; The optimized product recommendation sequence analyzes the style and material details of the products, generates highly attractive summary text, matches user preferences, and determines the final personalized traffic-driving content, including: For the products in the optimized product recommendation sequence, the product description text is parsed to extract the style and material details of the products, thereby obtaining trend keywords and tactile selling points; Based on the trend keywords and tactile selling points, and combined with preset scene association templates, highly attractive summary text is generated; If the matching degree between the highly attractive summary text and the user preference cluster is lower than the preset matching threshold, the scene association content and trendy words are adjusted, and the summary text is regenerated until the matching degree reaches the standard, and then determined as the final personalized traffic-driving content.

6. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the data processing method for a home furnishing industry lead generation platform as described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the data processing method for a home furnishing industry lead generation platform as described in any one of claims 1 to 5.

Images