Automatic recommendation marketing method and system for online figures

Abstract

The present invention discloses an automatic recommendation marketing method and system for online figures, the method comprising: collecting user behavior data on an online figure sales platform; constructing a user portrait according to the behavior data, and determining a preliminary recommendation list of figure products corresponding to the user portrait; obtaining the user's interest dimensions for each figure product involved in the behavior data, and calculating the user's emphasis on each interest dimension and the interest in each figure product; determining the user's ideal product feature vector according to the interest dimension, emphasis and interest; extracting each product feature vector in the preliminary recommendation list, calculating the similarity between the ideal product feature vector and each product feature vector, and screening the figure products that are finally used for recommendation marketing according to the similarity. By using the embodiments of the present invention, the recommendation strategy can be dynamically adjusted, the user characteristics can be accurately matched, and the personalized needs of the user can be better adapted, thereby improving the quality of recommendation and the user's satisfaction.

Description

Technical Field

[0001] The present invention belongs to the technical field of data mining recommendation, and in particular to an automatic recommendation marketing method and system for online figures. Background Art

[0002] In recent years, with the rapid development of e-commerce, the number of consumers who buy collectibles such as action figures online has gradually increased. As an important cultural derivative, action figures cover multiple fields such as animation and games, attracting a large number of fans and collectors. In order to meet the diverse needs of users, online action figure sales platforms are usually faced with the challenge of how to effectively recommend products that users may be interested in.

[0003] In traditional recommendation systems, commonly used methods are mainly based on collaborative filtering, content recommendation and other technologies. Although these methods can improve the accuracy of recommendations to a certain extent, they still have some shortcomings. For example, collaborative filtering methods rely on a large amount of user behavior data, and the recommendation effect for new users or cold start users is often poor. At the same time, content recommendation focuses on the characteristics of the items themselves and fails to fully consider the personalized needs and interest changes of users. In addition, traditional methods are usually unable to effectively capture the dynamic interests of users, resulting in a lack of targeted recommendations. Summary of the invention

[0004] The purpose of the present invention is to provide an automatic recommendation marketing method and system for online figures to address the deficiencies in the prior art, dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the personalized needs of users, thereby improving the quality of recommendations and user satisfaction.

[0005] An embodiment of the present application provides an automatic recommendation marketing method for online figures, the method comprising:

[0006] Collecting user behavior data on the online model sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, comment, and share behavior data;

[0007] Build a user profile based on the behavior data, and determine a preliminary recommended list of action figures corresponding to the user profile;

[0008] Obtaining the user's interest dimensions for each figurine product involved in the behavior data, and calculating the user's emphasis on each interest dimension and the user's interest in each figurine product;

[0009] Determining an ideal product feature vector of the user according to the interest dimension, the importance and the interest;

[0010] Extract each product feature vector in the preliminary recommendation list, calculate the similarity between the ideal product feature vector and each product feature vector, and screen the action figures that are ultimately used for recommendation marketing according to the similarity.

[0011] Optionally, the calculation formula of the importance is:

[0012] ;

[0013] Among them, the is the importance of the i-th interest dimension di, is the behavior frequency in the i-th interest dimension, is the weight coefficient of the i-th interest dimension, is the exponential coefficient of the i-th interest dimension, is the average behavior frequency of all interest dimensions, is the deviation parameter of the i-th interest dimension, and n is the total number of interest dimensions.

[0014] Optionally, the calculation formula of the interest degree is:

[0015] ;

[0016] Among them, the is the interest in the k-th figurine product, is the number of behaviors of the jth behavior for the kth figurine product, is the weight factor of the jth behavior, m is the number of behaviors for the kth figure product, Score the social influence of the k-th figurine product in the user's social circle. is the total number of user actions on all figurine products. Score the user's emotional evaluation of the k-th figurine product. is the emotional impact factor, is the time when the user last interacted with the k-th figurine product. is the time decay factor.

[0017] Optionally, determining the user's ideal product feature vector according to the interest dimension, the importance and the interest level includes:

[0018] Searching for a figurine product whose interest level is higher than a preset interest level from among the figurine products involved in the behavior data;

[0019] Extract the product partial feature vector corresponding to each interest dimension in the searched figurine product, weight each extracted product partial feature vector by the interest degree, and obtain the ideal product partial feature vector corresponding to each interest dimension;

[0020] The partial feature vectors of each ideal product are weighted and combined using the importance to obtain the feature vector of the ideal product that the user is most interested in.

[0021] Another embodiment of the present application provides an automatic recommendation and marketing system for online figures, the system comprising:

[0022] A collection module is used to collect the behavior data of users on the online figure sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, evaluate, and share behavior data;

[0023] A first determination module is used to construct a user profile based on the behavior data and determine a preliminary recommendation list of action figures corresponding to the user profile;

[0024] A calculation module, used to obtain the user's interest dimensions for each figurine product involved in the behavior data, and calculate the user's emphasis on each interest dimension and the user's interest in each figurine product;

[0025] A second determination module is used to determine the user's ideal product feature vector according to the interest dimension, the importance and the interest;

[0026] The recommendation module is used to extract the feature vectors of each product in the preliminary recommendation list, calculate the similarity between the feature vector of the ideal product and the feature vectors of each product, and screen the action figures that are ultimately used for recommendation marketing according to the similarity.

[0027] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0028] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0029] Compared with the prior art, the present invention provides an automatic recommendation marketing method for online figures, which collects user behavior data on an online figure sales platform; constructs a user portrait based on the behavior data, and determines a preliminary recommendation list of figure products corresponding to the user portrait; obtains the user's interest dimensions for each figure product involved in the behavior data, and calculates the user's emphasis on each interest dimension and the user's interest in each figure product; determines the user's ideal product feature vector based on the interest dimension, the emphasis and the interest; extracts each product feature vector in the preliminary recommendation list, calculates the similarity between the ideal product feature vector and each product feature vector, and screens the figure products ultimately used for recommendation marketing based on the similarity, thereby being able to dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the user's personalized needs, thereby improving the quality of recommendation and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A hardware structure block diagram of a computer terminal for an automatic online figure recommendation marketing method provided by an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of a process flow of an automatic recommendation marketing method for online figures provided by an embodiment of the present invention;

[0032] Figure 3 A schematic diagram of the structure of an automatic recommendation and marketing system for online figures provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0034] The embodiment of the present invention first provides an automatic recommendation marketing method for online figures, which can be applied to electronic devices such as computer terminals, specifically ordinary computers, etc.

[0035] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for an automatic online figure recommendation marketing method provided by an embodiment of the present invention. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the computer terminal may also include a transmission device 106 for communication functions and an input and output device 108. It can be understood by those skilled in the art that Figure 1The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0036] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / modules corresponding to the automatic recommendation marketing method of online figures in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0037] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0038] See also Figure 2 The embodiment of the present invention provides an automatic recommendation marketing method for online figures, which may include the following steps:

[0039] S201, collecting user behavior data on the online model sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, comment, and share behavior data;

[0040] The collection of behavioral data is the basis of the recommendation system, which aims to form a complete behavioral track of users by comprehensively and meticulously recording various operations of users on the online model sales platform. These behavioral data mainly include:

[0041] - Search behavior: Users enter keywords to find relevant action figures.

[0042] - Browsing behavior: User behavior of viewing product lists and details.

[0043] - Click behavior: The behavior of a user actually clicking on a product link or image.

[0044] - Add-to-cart behavior: The user adds a product to the shopping cart, indicating interest in the product.

[0045] - Collection behavior: Users collect products to show their long-term attention.

[0046] - Purchase behavior: Users complete transactions, which directly reflects the attractiveness of the product and the user's willingness to buy.

[0047] - Evaluation behavior: Users evaluate the purchased figures and provide feedback on their experience and product quality.

[0048] - Sharing behavior: Users share products or reviews with others, demonstrating social influence and recommendation intention.

[0049] Function and significance:

[0050] 1. Data-driven personalized recommendations: Complete behavioral data provides real data support for personalized recommendation algorithms, enabling the recommendation system to accurately match user preferences based on user historical behaviors and actual needs.

[0051] 2. User interest analysis: By analyzing the user's various behaviors, we can fully understand the user's interests and tendencies, helping the system to better understand the user's needs.

[0052] 3. Improve user experience: Real-time collection of user behavior can quickly respond to changes in user interests, thereby adjusting recommendation strategies in a timely manner and improving user shopping experience.

[0053] 4. Marketing strategy optimization: The aggregation and analysis of behavioral data can provide merchants with important market insights, helping them to develop more targeted marketing strategies, thereby improving conversion rates. One way to achieve this can include:

[0054] Step 1: Design of real-time behavioral data collection framework

[0055] - Event Listener Settings: In the online model sales platform, event listeners are used to monitor every user operation in real time to ensure comprehensive collection of various behavior data. These listeners can be set to capture click and input events of specific UI elements (such as search boxes, product icons, buttons, etc.).

[0056] Step 2: Behavioral data classification and time stamp recording

[0057] - Data classification module: During the process of collecting behavioral data, a data classification module is designed to classify the captured raw behavioral data by category to ensure structured storage of various behaviors (search, browse, click, etc.). In addition, each behavioral data needs to be timestamped to record the time of the user's operation for subsequent analysis.

[0058] Step 3: Data aggregation and storage

[0059] - Message queue system: Use message queues (such as Kafka) to process multiple events in parallel and transmit real-time collected user behavior data to the backend data warehouse. This design can ensure data stability in high concurrency situations and prevent data loss.

[0060] Step 4: Behavioral data preprocessing and feature extraction

[0061] - Data cleaning and preprocessing: In the backend data warehouse, data cleaning techniques (such as deduplication and noise filtering) are used to process the collected raw behavior data. Subsequently, feature extraction techniques (such as the frequent item set selection algorithm) are applied to extract and summarize the key features in the user behavior data and generate a structured user behavior feature vector.

[0062] Step 5: Dynamic update mechanism of behavioral data

[0063] - Incremental update algorithm: To ensure real-time performance, an incremental update algorithm is introduced. At regular intervals (e.g., ten minutes), the newly collected behavior data is merged with the existing data, and the user portrait and recommendation model are updated through a merging algorithm (such as weighted average, sliding window). This mechanism ensures that the recommendation model can reflect the latest changes in user interests.

[0064] Through the above steps, the online model sales platform can not only effectively collect user behavior data, but also ensure the timeliness and accuracy of the data, providing a solid foundation for subsequent user feature analysis and recommendation model optimization. This complex and dynamic implementation method enables the system to respond more flexibly to changes in user behavior, thereby improving the overall recommendation effect and user satisfaction.

[0065] S202, constructing a user profile based on the behavior data, and determining a preliminary recommendation list of action figures corresponding to the user profile;

[0066] User portraits are a comprehensive description of user preferences, interests, and potential needs by analyzing user behavior data. The process of building user portraits mainly includes the following aspects:

[0067] 1. Data integration: Integrate various types of user behavior data (such as search records, browsing history, add to cart, purchase, etc.) to form a panoramic view of the user.

[0068] 2. Feature extraction: Extract key features related to action figures from the integrated data, such as commonly searched product types, preferred brands, and commonly purchased price ranges.

[0069] 3. Portrait modeling: The extracted features are modeled through a specific algorithm to form a user portrait that can be used for recommendation, facilitating the use of subsequent recommendation engines.

[0070] 4. Initial recommendation generation: Based on the user portrait and the characteristic information in the figure product library, a preliminary recommendation list matching the user portrait is generated. For example, if the user portrait is a 2D otaku Japanese anime fan, a preliminary recommendation of 2D Japanese anime figures can be given; if the user portrait is a Chinese anime fan, a preliminary recommendation of popular Chinese anime series figures can be given, and so on. Function and significance:

[0071] 1. Improved personalized experience: Through detailed user portraits, the system can understand users' personalized needs and provide product recommendations that are more in line with their interests and preferences, thereby improving users' shopping experience.

[0072] 2. Precision marketing: Based on user portrait information, the platform can implement more precise marketing strategies and improve sales conversion rates by recommending action figures that are highly consistent with the user portrait.

[0073] 3. Dynamic adjustment: User portraits can be dynamically updated as user behavior changes, allowing the recommendation system to adapt to changes in user interests in a timely manner and ensure the relevance and accuracy of recommendations.

[0074] 4. Behavior analysis: The construction of user portraits not only helps with personalized recommendations, but also provides the platform with basic data for user behavior analysis, helping the platform to continuously improve its products and services. One implementation method may include:

[0075] Step 1: Data integration and preprocessing

[0076] - Raw data collection: Collect all user behavior data on the platform in real time, including key actions such as searching, browsing, and adding to cart, and store them in a temporary database for subsequent analysis.

[0077] - Data cleaning and filtering: Apply data cleaning algorithms to remove invalid, incomplete or duplicate behavior data, and then annotate the behavior data for subsequent analysis. Annotation can determine the weighting of each behavior in the user portrait based on its importance.

[0078] Step 2: Feature extraction and dimension division

[0079] - Behavioral feature analysis: Conduct in-depth analysis of user behavior data and use clustering algorithms (such as K-means) to divide user behavior features into different dimensions. For example, the user's preferred model types, brands, and price ranges can be extracted based on search keywords, browsing time, and purchase frequency.

[0080] - User preference induction: By calculating the commonality of each dimension, the basic preferences of users can be summarized. For example, users may prefer a certain series of figures or a specific price range.

[0081] Step 3: Portrait modeling

[0082] - Vectorized representation: vectorize the user's behavior features, for example, convert the feature values ​​of each dimension into a standardized vector through normalization. This process can use the TF-IDF (term frequency-inverse document frequency) model or Word2Vec and other technologies to facilitate subsequent calculations and comparisons.

[0083] - Multi-level user portrait construction: Establish multi-level user portraits, including basic portraits (such as gender, age, region, etc.) and behavioral portraits (such as interests, purchasing power, activity, etc.), to form a comprehensive, multi-dimensional user portrait.

[0084] Step 4: Preliminary recommendation list generation

[0085] - Feature matching algorithm: Use the feature information in the user portrait to calculate the similarity with the feature vector of each product in the figurine product library. You can use algorithms such as cosine similarity and Euclidean distance. The calculation will yield a list of figurine products that best match the user portrait.

[0086] - Threshold screening: Set an appropriate similarity threshold to screen out the action figures that meet the user's expectations and interests from the calculated candidate products, form a preliminary recommendation list, and prepare for subsequent optimization and verification.

[0087] Step 5: Dynamic Adjustment and Feedback Mechanism

[0088] - User behavior feedback monitoring: After the recommendation list is displayed, monitor the user's behavioral feedback on the recommended products (such as clicks, purchases, reviews, etc.), and feed it back into the update of the user portrait as new input data.

[0089] - Regularly update user portraits: set a period (such as daily or weekly) to regularly update user portraits to better reflect the latest changes in users' interests and needs. This dynamic update mechanism ensures that the recommendation system can respond to changes in user behavior in real time and improve the accuracy of recommendations.

[0090] Through the above steps, an accurate and dynamic user portrait is constructed, which not only lays a solid foundation for the generation of the preliminary recommendation list, but also provides effective support for subsequent accurate recommendations and personalized marketing. This complex implementation method enables the recommendation system to continuously adapt to changes in user needs, improve user satisfaction and the sales conversion rate of the platform.

[0091] S203, obtaining the user's interest dimensions for each figurine product involved in the behavior data, and calculating the user's emphasis on each interest dimension and the user's interest in each figurine product;

[0092] Extract interest dimensions related to action figures from user behavior data, such as brand preference, price range, action figure type (such as anime, games, movies, etc.). This step identifies the interest characteristics of users in these action figures by analyzing their search, browsing, and purchasing behaviors.

[0093] Through quantitative indicators such as behavior frequency, interest dimension weight, index coefficient, etc., the degree of importance users attach to each interest dimension is calculated, so that it can reflect the impact of different interest dimensions on users' decisions when choosing products.

[0094] Quantify the user's behavior data on different figurines (such as clicks, add-to-cart, social influence, emotional evaluation, etc.) and calculate the interest level of each figurine product. This process can help identify the user's attention to a specific figurine product.

[0095] 1. Enhance the accuracy of personalized recommendations: By analyzing the user's interest dimensions and importance, the system can more comprehensively understand the user's preferences and provide them with more desired action figure product recommendations.

[0096] 2. Improve user experience: By identifying users’ interest characteristics, we can create recommendation lists that are more in line with user needs, thereby improving users’ shopping experience and satisfaction.

[0097] 3. Promote the adjustment of marketing strategies: When analyzing user behavior, understanding the changes in interest dimensions and importance can help adjust marketing strategies so as to push the most attractive products to users at the right time and channel.

[0098] 4. Support dynamic recommendations: Users’ interests and priorities change dynamically. Real-time updates of these data can enable the recommendation system to quickly adapt to changes in user needs and achieve real-time personalized recommendations.

[0099] Specifically, a calculation formula for importance can be:

[0100] ;

[0101] This formula is used to calculate the importance that users attach to each interest dimension, reflecting the frequency and importance of users' behaviors on different interest dimensions.

[0102] Among them, the is the importance attached to the i-th interest dimension di, indicating the degree of influence of this dimension on the user's decision. is the behavior frequency in the i-th interest dimension, reflecting the user's activity intensity in this dimension. is the weight coefficient of the i-th interest dimension, indicating the importance of this dimension in the opinion of experts in the relevant field, which can be obtained through previous user behavior analysis. is the exponential coefficient of the i-th interest dimension, which can be optimized according to the data distribution or set according to the prior data, and is usually used to emphasize the impact of user behavior. is the average behavior frequency of all interest dimensions, used to standardize the influence of different dimensions. is the deviation parameter of the ith interest dimension, controlling the response curve of the importance to the behavior frequency, which can be obtained by data fitting or set according to prior data. The n is the total number of interest dimensions, reflecting the range of the multiple interest dimensions concerned.

[0103] Specifically, a calculation formula for interest degree can be:

[0104] ;

[0105] This formula is used to calculate the user's interest in the kth figurine product. It combines the user's behavior, emotional evaluation and time factors, and ultimately reflects the user's attention to the figurine.

[0106] Among them, the is the interest in the kth figurine product, indicating the user’s attention intensity to the product. is the number of actions of the jth behavior for the kth figurine product, and records the different interactive behaviors of users on the product. is the weight factor of the jth behavior, reflecting the influence of each behavior on the final interest level, which can be set according to the importance of the user behavior. The social influence score of the k-th figurine product in the user's social circle reflects the impact of social sharing factors on product interest. The quantitative score can be performed by the user. is the total number of user actions on all figurine products. The user's emotional evaluation score for the kth figurine product conveys the user's positive or negative emotions towards the product, and the user can make a quantitative score by himself. is the emotional influencing factor, which determines the influence of emotional evaluation on interest. The last time the user interacted with the k-th figurine product is used to reduce the interest of users who have not interacted for a long time, so that the interest of users who have not interacted for a long time can be reasonably weakened, while maintaining the interest of users who have recently interacted. It is a time decay factor, which is used to control the impact of time on interest and reflects the user's time sensitivity.

[0107] The design of the formula combines user behavior, emotion and social influence to form a multi-dimensional interest evaluation system; the emotional evaluation score is introduced to quantify the user's subjective emotional response to the product in order to more delicately reflect the user's real interest; the social influence introduces the factor of social network, taking into account the influence of the product in the social circle, making popular products more attractive; the time decay factor increases the weight of the latest interaction by processing the logarithm of time, while reducing the impact of the interaction over a long time on the interest.

[0108] S204, determining an ideal product feature vector of the user according to the interest dimension, the importance and the interest;

[0109] This step creates an "ideal product feature vector" that reflects user preferences by combining the interest dimensions, importance, and interest in user behavior data. This work helps to deeply personalize recommendations, thereby improving user experience and purchase conversion rates. Specifically:

[0110] - The interest dimension indicates the user's preference for different attributes or characteristics. These dimensions may include price, brand, style, material, etc., reflecting the user's focus when choosing figures.

[0111] - Importance refers to the degree of attention paid by users to various interest dimensions. The calculated importance can better reflect the strength of users' preferences for specific attributes.

[0112] - Interest level refers to the actual level of interest a user has in a specific action figure product. By analyzing user behavior data (such as browsing, purchasing, etc.), we can measure the user's attraction to a specific product.

[0113] Combining the above three factors, the determination of the ideal product feature vector can help the platform to recommend action figures that suit the user's taste, thereby improving user satisfaction and the platform's sales performance. By determining the user's ideal product feature vector, the user's multi-dimensional interests and preferences can be effectively quantified, thereby providing an efficient and more targeted basis for subsequent product recommendations. This can not only improve user satisfaction, but also increase purchase conversion rates and improve the overall benefits of the platform. At the same time, dynamically adjusting the feature vector can ensure that the recommendation system continues to adapt to changes in user behavior and enhance user stickiness.

[0114] Specifically, the figurine products with the interest level higher than the preset interest level may be searched from the figurine products involved in the behavior data;

[0115] The purpose of this step is to filter out the products that users have clearly shown strong interest in from the existing figures. By setting a threshold, the system can effectively filter out figures that users may not be interested in, ensuring that the basic data for subsequent analysis is what users are actually interested in. This screening process can effectively reduce the amount of calculation, improve the relevance of recommendations, and ensure that the product feature vectors used in subsequent steps can reflect the user's actual preferences.

[0116] Extract the product partial feature vector corresponding to each interest dimension in the searched figurine product, weight each extracted product partial feature vector by the interest degree, and obtain the ideal product partial feature vector corresponding to each interest dimension;

[0117] In this process, the system extracts the feature vectors of each interest dimension from the selected figures. By weighting these feature vectors (the weight is the interest level), the feature vectors of the ideal product for each interest dimension are calculated. This is based on the user's emphasis on different interest dimensions and the actual products of interest, in order to form a feature vector that better meets the user's expectations. This step ensures that the user's preferences are accurately captured and play a role in subsequent recommendations, enhancing the effectiveness of personalized recommendations.

[0118] For example, if 3 figurines are found, the interest level is , the weight value for quantifying interest is Each figurine product is assumed to have three interest dimensions, and the first interest dimension a corresponds to the three partial feature vectors of the figurine products. , then the ideal product partial feature vector corresponding to the first interest dimension a is: , similarly, we can get the ideal product feature vectors of other interest dimensions.

[0119] The partial feature vectors of each ideal product are weighted and combined using the importance to obtain the feature vector of the ideal product that the user is most interested in.

[0120] In this step, the system will perform a weighted combination of the ideal product partial feature vectors of each interest dimension obtained previously, with the weight being the user's emphasis on each interest dimension, that is, the ideal product partial feature vector of each interest dimension is multiplied by the quantitative weight corresponding to the emphasis, and then summed to obtain the ideal product feature vector. The ideal product feature vector finally generated will be multi-dimensional and can fully reflect the user's preferences. This final feature vector represents the product features that the user is most interested in, and can provide a strong data foundation for subsequent recommendation algorithms, thereby improving the accuracy and satisfaction of recommendations.

[0121] Through the above methods, the platform can achieve accurate personalized recommendations and improve user satisfaction and engagement.

[0122] S205, extracting the feature vectors of each product in the preliminary recommendation list, calculating the similarity between the feature vector of the ideal product and the feature vectors of each product, and screening the action figures to be finally used for recommendation marketing according to the similarity.

[0123] The key to this step is to use the user's behavioral data and the ideal product feature vector to focus on the action figures in the preliminary recommendation list, and finally determine which products are more in line with the user's preferences. First, the product feature vector is extracted to establish a method that can quantify product characteristics so that subsequent similarity calculations can be performed in the same dimension. This step ensures that the recommendation system can not only take into account the user's past preferences, but also convert them into specific product features to facilitate accurate comparison of the similarities between different products. Function and significance:

[0124] 1. Accurate recommendation: By calculating the similarity between the ideal product feature vector and the candidate products, we can accurately identify the products that best meet the user's needs. This method effectively improves the relevance and efficiency of recommendations.

[0125] 2. Personalized experience: Based on the user's unique interest dimensions and importance, a user-specific product feature vector is constructed so that the recommendation system can truly meet the preferences of individual users and enhance the user's shopping experience.

[0126] 3. Data-driven decision-making: Through similarity calculation based on data analysis, the influence of subjective bias is reduced to ensure the scientificity and rationality of recommendation decisions. One implementation method may include:

[0127] 1. Feature vector extraction:

[0128] - Define feature dimensions: First, define multiple feature dimensions for each model product in the preliminary recommendation list, such as brand, style, price, material, user reviews, etc. The feature dimensions need to be consistent with the feature vector dimensions of the ideal product.

[0129] - Construct a feature matrix: Construct a feature vector for each figurine product. For example, you can use a numerical method to encode the "style" with different numerical values ​​(such as 1 for "anime style" and 2 for "realistic style") and generate a product feature vector.

[0130] 2. Calculate similarity:

[0131] - Choose a similarity metric: Use methods such as cosine similarity or Euclidean distance to measure the similarity between the feature vector of the ideal product and the feature vectors of each product in the initial recommendation.

[0132] - Perform similarity calculation: Substitute the ideal product feature vector and each product feature vector into the similarity calculation formula, such as cosine similarity, etc.

[0133] 3. Sorting and filtering results:

[0134] - Generate similarity list: Based on the calculated similarities, generate a similarity score list for each candidate product.

[0135] - Set a threshold: Set a similarity threshold and filter out products with similarity higher than the threshold as the final recommended products.

[0136] - Sorting recommendation: Sort the remaining products according to the similarity value, give priority to displaying the products with the highest similarity, and form the final recommendation list.

[0137] Through the above processing flow, this method ensures a high degree of consistency between recommended products and user preferences, greatly improving the effect of online figure marketing.

[0138] It can be seen that the user's behavioral data on the online figure sales platform is collected; a user profile is constructed based on the behavioral data, and a preliminary recommendation list of figure products corresponding to the user profile is determined; the user's interest dimensions for each figure product involved in the behavioral data are obtained, and the user's emphasis on each interest dimension and the user's interest in each figure product are calculated; the user's ideal product feature vector is determined based on the interest dimension, the emphasis and the interest; the feature vectors of each product in the preliminary recommendation list are extracted, and the similarity between the ideal product feature vector and each product feature vector is calculated, and the figure products ultimately used for recommendation marketing are screened based on the similarity, so as to dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the user's personalized needs, thereby improving the quality of recommendation and user satisfaction.

[0139] Another embodiment of the present invention provides an automatic recommendation and marketing system for online figures, see Figure 3 , the system may include:

[0140] The collection module 301 is used to collect the behavior data of users on the online figure sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, evaluate, and share behavior data;

[0141] A first determination module 302 is used to construct a user profile based on the behavior data and determine a preliminary recommendation list of action figures corresponding to the user profile;

[0142] The calculation module 303 is used to obtain the user's interest dimension for each figure product involved in the behavior data, and calculate the user's emphasis on each interest dimension and the user's interest in each figure product;

[0143] A second determination module 304 is used to determine the user's ideal product feature vector according to the interest dimension, the importance and the interest;

[0144] The recommendation module 305 is used to extract the feature vectors of each product in the preliminary recommendation list, calculate the similarity between the feature vector of the ideal product and the feature vectors of each product, and screen the action figures to be recommended for marketing according to the similarity.

[0145] It can be seen that the user's behavioral data on the online figure sales platform is collected; a user profile is constructed based on the behavioral data, and a preliminary recommendation list of figure products corresponding to the user profile is determined; the user's interest dimensions for each figure product involved in the behavioral data are obtained, and the user's emphasis on each interest dimension and the user's interest in each figure product are calculated; the user's ideal product feature vector is determined based on the interest dimension, the emphasis and the interest; the feature vectors of each product in the preliminary recommendation list are extracted, and the similarity between the ideal product feature vector and each product feature vector is calculated, and the figure products ultimately used for recommendation marketing are screened based on the similarity, so as to dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the user's personalized needs, thereby improving the quality of recommendation and user satisfaction.

[0146] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0147] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps:

[0148] S201, collecting user behavior data on the online model sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, comment, and share behavior data;

[0149] S202, constructing a user profile based on the behavior data, and determining a preliminary recommendation list of action figures corresponding to the user profile;

[0150] S203, obtaining the user's interest dimensions for each figurine product involved in the behavior data, and calculating the user's emphasis on each interest dimension and the user's interest in each figurine product;

[0151] S204, determining an ideal product feature vector of the user according to the interest dimension, the importance and the interest;

[0152] S205, extracting the feature vectors of each product in the preliminary recommendation list, calculating the similarity between the feature vector of the ideal product and the feature vectors of each product, and screening the action figures to be finally used for recommendation marketing according to the similarity.

[0153] It can be seen that the user's behavioral data on the online figure sales platform is collected; a user profile is constructed based on the behavioral data, and a preliminary recommendation list of figure products corresponding to the user profile is determined; the user's interest dimensions for each figure product involved in the behavioral data are obtained, and the user's emphasis on each interest dimension and the user's interest in each figure product are calculated; the user's ideal product feature vector is determined based on the interest dimension, the emphasis and the interest; the feature vectors of each product in the preliminary recommendation list are extracted, and the similarity between the ideal product feature vector and each product feature vector is calculated, and the figure products ultimately used for recommendation marketing are screened based on the similarity, so as to dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the user's personalized needs, thereby improving the quality of recommendation and user satisfaction.

[0154] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0155] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0156] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0157] S201, collecting user behavior data on the online model sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, comment, and share behavior data;

[0158] S202, constructing a user profile based on the behavior data, and determining a preliminary recommendation list of action figures corresponding to the user profile;

[0159] S203, obtaining the user's interest dimensions for each figurine product involved in the behavior data, and calculating the user's emphasis on each interest dimension and the user's interest in each figurine product;

[0160] S204, determining an ideal product feature vector of the user according to the interest dimension, the importance and the interest;

[0161] S205, extracting the feature vectors of each product in the preliminary recommendation list, calculating the similarity between the feature vector of the ideal product and the feature vectors of each product, and screening the action figures to be finally used for recommendation marketing according to the similarity.

[0162] It can be seen that the user's behavioral data on the online figure sales platform is collected; a user profile is constructed based on the behavioral data, and a preliminary recommendation list of figure products corresponding to the user profile is determined; the user's interest dimensions for each figure product involved in the behavioral data are obtained, and the user's emphasis on each interest dimension and the user's interest in each figure product are calculated; the user's ideal product feature vector is determined based on the interest dimension, the emphasis and the interest; the feature vectors of each product in the preliminary recommendation list are extracted, and the similarity between the ideal product feature vector and each product feature vector is calculated, and the figure products ultimately used for recommendation marketing are screened based on the similarity, so as to dynamically adjust the recommendation strategy, accurately match user characteristics, and better adapt to the user's personalized needs, thereby improving the quality of recommendation and user satisfaction.

[0163] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. An automatic recommendation marketing method for online figures, characterized in that: The method comprises: Collecting user behavior data on the online model sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, comment, and share behavior data; Build a user profile based on the behavior data, and determine a preliminary recommended list of action figures corresponding to the user profile; wherein all the user's behavior data on the platform is collected in real time, and all the behavior data is integrated and pre-processed; Analyze the preprocessed behavior data, use clustering algorithms to divide the user's behavior characteristics into different dimensions, and convert the feature values ​​of each dimension into a standardized vector through normalization processing to establish a multi-dimensional user portrait, which includes a basic portrait and a behavior portrait; Using the feature information in the user portrait, similarity calculation is performed with the feature vector of each product in the figurine product library, and a preliminary recommendation list of figurine products corresponding to the user portrait is obtained through threshold screening; Obtain the user's interest dimensions for each figure product involved in the behavior data, and calculate the user's emphasis on each interest dimension and the user's interest in each figure product; wherein the interest dimension related to the figure product is extracted from the user's behavior data; the calculation formula for the emphasis is: Among them, the is the importance of the i-th interest dimension di, is the behavior frequency in the i-th interest dimension, is the weight coefficient of the i-th interest dimension, is the exponential coefficient of the i-th interest dimension, is the average behavior frequency of all interest dimensions, is the deviation parameter of the i-th interest dimension, and n is the total number of interest dimensions; the calculation formula of the interest degree is: Among them, the is the interest in the k-th figurine product, is the number of behaviors of the jth behavior for the kth figurine product, is the weight factor of the jth behavior, m is the number of behaviors for the kth figure product, Score the social influence of the k-th figurine product in the user's social circle. is the total number of user actions on all figurine products. Score the user's emotional evaluation of the k-th figurine product. is the emotional impact factor, is the time when the user last interacted with the k-th figurine product. is the time decay factor; Determine the user's ideal product feature vector according to the interest dimension, the importance and the interest level; wherein, from each figurine product involved in the behavior data, search for a figurine product whose interest level is higher than a preset interest level; Extract the product partial feature vector corresponding to each interest dimension in the searched figurine product, weight each extracted product partial feature vector by the interest degree, and obtain the ideal product partial feature vector corresponding to each interest dimension; The partial feature vectors of each ideal product are weighted and combined by using the importance to obtain the feature vector of the ideal product that the user is most interested in; Extract each product feature vector in the preliminary recommendation list, calculate the similarity between the ideal product feature vector and each product feature vector, and screen the action figures that are ultimately used for recommendation marketing according to the similarity.

2. An automatic recommendation marketing system for online figures, characterized in that: The system comprises: A collection module is used to collect the behavior data of users on the online figure sales platform, wherein the behavior data includes: search, browse, click, add to cart, favorite, purchase, evaluate, and share behavior data; The first determination module is used to construct a user profile based on the behavior data and determine a preliminary recommendation list of action figures corresponding to the user profile; wherein all the user's behavior data on the platform is collected in real time, and all the behavior data is integrated and pre-processed; Analyze the preprocessed behavior data, use clustering algorithms to divide the user's behavior characteristics into different dimensions, and convert the feature values ​​of each dimension into a standardized vector through normalization processing to establish a multi-dimensional user portrait, which includes a basic portrait and a behavior portrait; Using the feature information in the user portrait, similarity calculation is performed with the feature vector of each product in the figurine product library, and a preliminary recommendation list of figurine products corresponding to the user portrait is obtained through threshold screening; A calculation module is used to obtain the user's interest dimensions for each figure product involved in the behavior data, and calculate the user's emphasis on each interest dimension and the user's interest in each figure product; wherein the interest dimension related to the figure product is extracted from the user's behavior data; the calculation formula for the emphasis is: Among them, the is the importance of the i-th interest dimension di, is the behavior frequency in the i-th interest dimension, is the weight coefficient of the i-th interest dimension, is the exponential coefficient of the i-th interest dimension, is the average behavior frequency of all interest dimensions, is the deviation parameter of the i-th interest dimension, and n is the total number of interest dimensions; the calculation formula of the interest degree is: Among them, the is the interest in the k-th figurine product, is the number of behaviors of the jth behavior for the kth figurine product, is the weight factor of the jth behavior, m is the number of behaviors for the kth figure product, Score the social influence of the k-th figurine product in the user's social circle. is the total number of user actions on all figurine products. Score the user's emotional evaluation of the k-th figurine product. is the emotional impact factor, is the time when the user last interacted with the k-th figurine product. is the time decay factor; A second determination module is used to determine the user's ideal product feature vector according to the interest dimension, the importance and the interest level; wherein, from each figurine product involved in the behavior data, a figurine product whose interest level is higher than a preset interest level is searched; Extract the product partial feature vector corresponding to each interest dimension in the searched figurine product, weight each extracted product partial feature vector by the interest degree, and obtain the ideal product partial feature vector corresponding to each interest dimension; The partial feature vectors of each ideal product are weighted and combined by using the importance to obtain the feature vector of the ideal product that the user is most interested in; The recommendation module is used to extract the feature vectors of each product in the preliminary recommendation list, calculate the similarity between the feature vector of the ideal product and the feature vectors of each product, and screen the action figures that are ultimately used for recommendation marketing according to the similarity.

3. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method of claim 1 when executed.

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

Images