Marketing decision-making method and system for e-commerce platform

Abstract

The present invention discloses a marketing decision method and system for an e-commerce platform, which belongs to the field of Internet technology, wherein the method includes: sending a data collection instruction to a user terminal, collecting the user's browsing behavior data and transaction data at the user terminal to form a user original data set; constructing a user feature vector, performing cluster analysis on the user feature vector, and dividing the user into a regular user group, a high-value common user group, and a high-value individual user group; generating a group interaction strategy template for each user group; combining the group interaction strategy template and the user's original data set at the user terminal to generate a user personalized interaction strategy. This application solves the problem of lack of personalization and precision in the marketing decision of the e-commerce platform in the prior art, and it is difficult to meet the differentiated needs of different user groups, and achieves the technical effect of formulating personalized interaction strategies for different user groups, improving the accuracy and effectiveness of marketing decisions, and meeting the differentiated needs of different users.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a marketing decision-making method and system for an e-commerce platform. Background Art

[0002] With the rapid development and widespread application of Internet technology, e-commerce has become an important part of modern business activities. More and more companies conduct business and establish connections with customers through e-commerce platforms. However, in the existing technology, e-commerce platforms adopt a unified marketing strategy to push the same marketing content to all users, ignoring the individual differences and diversified needs of users, lacking personalization and precision, and unable to meet the differentiated needs of different user groups, resulting in poor marketing effects and poor user experience, which seriously affects user satisfaction. Summary of the invention

[0003] This application provides a marketing decision-making method and system for e-commerce platforms, aiming to solve the problem in the prior art that marketing decisions of e-commerce platforms lack personalization and precision, and it is difficult to meet the differentiated needs of different user groups.

[0004] In view of the above problems, the present application provides a marketing decision-making method and system for an e-commerce platform.

[0005] The first aspect disclosed in the present application provides a marketing decision-making method for an e-commerce platform, the method comprising: sending a data collection instruction to a user terminal, collecting the user's browsing behavior data and transaction data at the user terminal to form a user original data set; receiving the user original data set sent by the user terminal, performing data cleaning and feature extraction, and constructing a user feature vector containing a user perceived uniqueness score; performing cluster analysis on the user feature vector, dividing the users into a regular user group, a high-value common user group, and a high-value individual user group, wherein the user perceived uniqueness score of the high-value individual user group is higher than that of the high-value common user group; generating corresponding group interaction strategy templates for the regular user group, the high-value common user group, and the high-value individual user group, which are respectively a regular interaction strategy template, a high-value common interaction strategy template, and a high-value individual interaction strategy template; sending the corresponding group interaction strategy template to the user terminal, and generating a user personalized interaction strategy by combining the group interaction strategy template and the user's original data set at the user terminal.

[0006] Another aspect disclosed in the present application provides a marketing decision system for an e-commerce platform, the system comprising: a data collection module, used to send a data collection instruction to a user terminal, collect the user's browsing behavior data and transaction data at the user terminal, and form a user original data set; a data processing module, used to receive the user original data set sent by the user terminal, perform data cleaning and feature extraction, and construct a user feature vector containing a user perceived uniqueness score; a user classification module, used to perform cluster analysis on the user feature vector, and divide the users into a regular user group, a high-value common user group, and a high-value individual user group, wherein the user perceived uniqueness score of the high-value individual user group is higher than that of the high-value common user group; a strategy template generation module, used to generate corresponding group interaction strategy templates for the regular user group, the high-value common user group, and the high-value individual user group, respectively, which are a regular interaction strategy template, a high-value common interaction strategy template, and a high-value individual interaction strategy template; a user strategy generation module, used to send the corresponding group interaction strategy template to the user terminal, and generate a user personalized interaction strategy by combining the group interaction strategy template and the user's original data set at the user terminal.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] By sending data collection instructions to user terminals, the user's browsing behavior data and transaction data are collected at the user terminal to form a user original data set, and the user's behavior data on the e-commerce platform is obtained comprehensively and accurately, laying a data foundation for subsequent personalized interaction strategies; the user original data set sent by the user terminal is received, and data cleaning and feature extraction are performed to construct a user feature vector containing the user's perceived uniqueness score, providing a basis for subsequent user segmentation; cluster analysis is performed on the user feature vector to divide the users into regular user groups, high-value common user groups and high-value individual user groups, among which the user-perceived uniqueness score of the high-value individual user group is higher than that of the high-value common user group, providing support for the formulation of differentiated marketing strategies; corresponding group interaction strategy templates are generated for the regular user group, the high-value common user group and the high-value individual user group, respectively, namely the regular Interaction strategy templates, high-value common interaction strategy templates, and high-value personalized interaction strategy templates are designed according to the characteristics of different user groups to achieve personalization and differentiation of marketing decisions; the corresponding group interaction strategy templates are sent to the user terminal, and the group interaction strategy templates and the user's original data set are combined at the user terminal to generate a user personalized interaction strategy, which combines the group-level interaction strategy templates with the original data of the individual user to generate a personalized interaction strategy for each user, and realizes the refinement and precision of marketing decisions. This technical solution solves the problem of lack of personalization and precision in the marketing decisions of the e-commerce platform in the prior art, which makes it difficult to meet the differentiated needs of different user groups, and achieves the technical effect of formulating personalized interaction strategies for different user groups, improving the accuracy and effectiveness of marketing decisions, and meeting the differentiated needs of different users.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of a marketing decision-making method for an e-commerce platform is provided for an embodiment of the present application;

[0011] Figure 2 A structural diagram of a marketing decision-making system for an e-commerce platform is provided for an embodiment of the present application.

[0012] Explanation of the reference numerals: data collection module 11 , data processing module 12 , user classification module 13 , policy template generation module 14 , user policy generation module 15 . DETAILED DESCRIPTION

[0013] The overall idea of ​​the technical solution provided by this application is as follows:

[0014] The embodiments of the present application provide a marketing decision-making method and system for an e-commerce platform. First, by collecting browsing behavior data and transaction data on the user terminal, a comprehensive and real user original data set is formed. Then, the original data is cleaned and feature extracted to construct a user feature vector containing the user perceived uniqueness score. Furthermore, cluster analysis is used to group the user feature vectors, and regular user groups, high-value common user groups, and high-value individual user groups are distinguished according to the user perceived uniqueness score. Then, for different user groups, corresponding group interaction strategy templates are designed to reflect differentiated marketing strategies. Afterwards, the group interaction strategy template is combined with the user's original data to generate a personalized interaction strategy for each user.

[0015] In short, this application greatly improves the scientificity, accuracy and effectiveness of marketing decisions through user data collection, user segmentation, differentiated strategy formulation, and personalized strategy generation, meets the differentiated needs of different user groups, and provides strong support for the refined operation of e-commerce platforms and the improvement of user experience.

[0016] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0017] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a marketing decision method for an e-commerce platform, the method comprising:

[0018] S1: Send data collection instructions to the user terminal, collect the user's browsing behavior data and transaction data at the user terminal, and form the user's original data set.

[0019] Specifically, by sending data collection instructions to the terminal devices used by users (such as smart phones, tablets, personal computers, etc.), the user terminals are triggered and guided to start collecting data. After receiving the data collection instructions, the user terminals collect various types of data of users on the e-commerce platform in a targeted manner according to the content of the instructions, including browsing behavior data and transaction data. Among them, browsing behavior data is a detailed record of users browsing goods, stores, promotional activities, etc. on the e-commerce platform, such as the categories of goods browsed, price ranges, browsing depth, length of stay, etc.; transaction data is a detailed record of users placing orders to purchase goods on the e-commerce platform, such as the categories of goods purchased, quantity, amount, purchase frequency, payment method, etc.

[0020] After the user terminal collects browsing behavior data and transaction data, it stores and organizes them according to the format requirements in the data collection instructions, such as json, xml, etc., to form a structured user original data set, which covers all browsing and transaction behaviors of users within a certain time range and is the data basis for user analysis and marketing decisions. Among them, data collection is continuous and can dynamically update the user original data set to keep it synchronized with the user's latest behavior, providing the most timely and accurate behavior reference for marketing decisions.

[0021] S2: Receive the original user data set sent by the user terminal, perform data cleaning and feature extraction, and construct a user feature vector including the user perceived uniqueness score.

[0022] Specifically, after obtaining the original user data set, data cleaning and feature extraction are performed on it in order to convert it into a structured user feature vector suitable for cluster analysis. Data cleaning refers to the identification and processing of outliers, missing values, duplicate values, etc. in the original user data set to improve data quality. After completing data cleaning, key features that help characterize the user are extracted from the cleaned original user data set, including basic user characteristics, such as gender, age, region and other basic attributes; user behavior characteristics, such as browsing product categories, browsing time, purchase frequency, consumption amount and other behavioral attributes; user preference characteristics, such as preferred product style, reaction to promotional activities and other interest attributes; user interaction characteristics, such as participation in product evaluation, order sharing frequency, and interaction data with platform employees (customer service, sales, etc.).

[0023] At the same time, the user's perceived uniqueness level has a significant impact on his response to marketing activities. Therefore, the user's perceived uniqueness score is obtained through the user's key characteristics to measure the degree to which the user perceives himself as different and unique from other users. The perceived uniqueness score is obtained by mining the user's key characteristics, analyzing the frequency of users purchasing unique or niche products, the density of expressing personalized demands in interactions, and the common grounding behavior in the process of interaction with users. Among them, the common grounding behavior is the behavior of consciously showing similarities with users in certain aspects during the interaction with users, thereby narrowing the psychological distance with users. Incorporating the perceived uniqueness score into the user feature vector, the personalized needs and marketing response tendencies of different users can be obtained, providing a basis for subsequent targeted decision-making.

[0024] S3: Perform cluster analysis on the user feature vectors to divide the users into a regular user group, a high-value common user group and a high-value individual user group, wherein the user perceived uniqueness score of the high-value individual user group is higher than that of the high-value common user group.

[0025] Specifically, after obtaining the user feature vector containing the user's perceived uniqueness score, the users are segmented and classified to identify user groups with different characteristics and needs, laying the foundation for subsequent personalized marketing decisions. Using the cluster analysis method, users are divided into several groups based on their similarities in feature dimensions. The characteristics of members within each group are similar, while the characteristics between groups vary greatly.

[0026] First, three user groups are preset, namely, the regular user group, the high-value common user group and the high-value individual user group. Among them, the regular user group refers to ordinary users who perform mediocrely in terms of purchasing power, loyalty, interaction frequency and other dimensions. These users have less interest and investment in brands and products, only basic shopping needs, and low requirements for personalized services; the high-value common user group refers to high-quality users who perform outstandingly in terms of purchasing power, loyalty, interaction frequency and other dimensions, but have low perceived uniqueness scores; the high-value individual user group refers to high-quality users who perform outstandingly in terms of purchasing power, loyalty, interaction frequency and other dimensions, and have high perceived uniqueness scores. In actual operation, according to the key indicators in the user feature vector, through algorithms such as K-means and hierarchical clustering, users are divided into several clusters, each cluster corresponds to a preliminary user group; then, according to the clustering results and the preset three user groups, the preliminary user groups are mapped, merged and adjusted to obtain the regular user group, the high-value common user group and the high-value individual user group.

[0027] S4: Generate corresponding group interaction strategy templates for the regular user group, the high-value common user group, and the high-value personalized user group, which are respectively a regular interaction strategy template, a high-value common interaction strategy template, and a high-value personalized interaction strategy template.

[0028] Specifically, after dividing users into regular user groups, high-value common user groups and high-value individual user groups through cluster analysis, differentiated group interaction strategy templates are designed according to the characteristics and needs of different groups to enhance customer experience and marketing effectiveness.

[0029] For the conventional user group, this part of users has low spending power and interaction frequency, and does not have high requirements for personalization. The interaction strategy is based on the principle of optimal cost-effectiveness, and adopts a unified and standardized marketing mix, such as regular promotions, new product recommendations, etc., to meet basic shopping needs, thus obtaining a conventional interaction strategy template. For the high-value common user group, this part of users has strong spending power and high loyalty, but does not have strong requirements for personalization. In the interaction, the method of perceived similarity (similar to users in attitudes, values, interests, etc.) or accidental similarity (similar to users in random interaction characteristics) is used to increase the occurrence of general common grounding behaviors, thus obtaining a high-value common interaction strategy template. For the high-value individual user group, this part of users has strong spending power, high loyalty, and strong requirements for personalization. Therefore, the unique personality of the user is deeply matched, and customized product recommendations and exclusive services are provided, thus obtaining a high-value individual interaction strategy template.

[0030] S5: Sending a corresponding group interaction strategy template to a user terminal, and combining the group interaction strategy template with the user original data set at the user terminal to generate a user personalized interaction strategy.

[0031] Specifically, after generating the general interaction strategy template, the high-value common interaction strategy template and the high-value individual interaction strategy template, we further generate personalized interaction strategies based on the group strategy according to the characteristics of individual users, so as to achieve more accurate and appropriate user reach and services.

[0032] First, the corresponding group interaction strategy template is sent to the user terminals belonging to different user groups. The user terminals of the conventional user group receive the conventional interaction strategy template, the user terminals of the high-value commonality user group receive the high-value commonality interaction strategy template, and the user terminals of the high-value individuality user group receive the high-value individuality interaction strategy template. Then, the user terminal obtains the user's original data set from the local storage, including the user's detailed interaction records, browsing history, purchase records, etc. By analyzing the user's original data set, key information that can reflect the individual characteristics of the user is extracted, such as the recently browsed goods, purchase cycle, coupon usage preference, etc., which reflects the user's unique personality beyond the overall characteristics of the group. Next, the user terminal matches and fuses the extracted user personalized information with the group interaction strategy template. Taking the high-value individual user as an example, the group strategy template has designed personalized recommendations, exclusive services and other strategies based on the high-perception uniqueness characteristics of the group. On this basis, the user terminal adjusts the recommended content according to the specific goods recently browsed by the user; selects the best interaction time according to the user's purchase time pattern; and matches the exclusive coupon according to the user's preferential preferences. By finely fusing the group strategy with the individual information, a user personalized interaction strategy that fully matches the unique characteristics of the user is generated.

[0033] By integrating group strategy templates and user original data sets, while following the overall interaction principle of the group, accurate reach of each user is achieved through the adjustment of personalized information, which not only saves the time and cost of formulating personalized strategies, but also improves the degree of personalization and matching.

[0034] Furthermore, the embodiment of the present application also includes:

[0035] After data cleaning and feature extraction of the original user data set, user behavior features, user preference features and user interaction features are obtained;

[0036] Obtaining a user perceived uniqueness score according to the user behavior characteristics, the user preference characteristics, and the user interaction characteristics;

[0037] The user behavior characteristics, the user preference characteristics, the user interaction characteristics and the user perceived uniqueness score are aggregated to generate the user feature vector.

[0038] In a feasible implementation, after data cleaning and feature extraction of the user's original data set, the corresponding user key features are extracted from it, including user behavior features, user preference features and user interaction features. Among them, the user behavior features reflect the user's actual use behavior on the platform, such as the category, frequency, and duration of browsing goods, the frequency and amount of purchasing goods, etc., which are obtained by statistical analysis of user behavior logs; user preference features reflect the user's personalized preferences for goods and services, such as the preferred product style, price range, brand, etc., which are obtained by mining and analyzing the user's collection records and purchase records; user interaction features reflect the user's interaction with the platform and merchants, such as the frequency of consulting customer service, the number of times participating in promotional activities, the frequency of evaluating goods, etc., which are obtained by statistical analysis of interaction logs. Then, machine learning modeling is used to train an evaluation model with the characteristics and perceived uniqueness scores of historical users as training data. The key features of the current user are input into the model to obtain its perceived uniqueness score. Afterwards, a vectorization method is used to map various features and perceived uniqueness scores to a high-dimensional vector space to form a user feature vector.

[0039] Furthermore, the embodiment of the present application also includes:

[0040] A historical user database of the interactive e-commerce platform, based on which a sample user feature set and a sample perceived uniqueness score set are constructed, wherein the sample user features include sample user behavior features, sample user preference features, and sample user interaction features;

[0041] Constructing a perceptual uniqueness evaluator based on the sample user feature set and the sample perceptual uniqueness score set;

[0042] The user behavior characteristics, the user preference characteristics and the user interaction characteristics are input into a perceived uniqueness evaluator to obtain a user perceived uniqueness score.

[0043] In a feasible implementation, a group of representative historical user samples are obtained from the historical user database of the e-commerce platform. These historical sample users cover various sub-groups and have different behavior patterns, preference characteristics and interaction situations. Then, for each sample user, extract its behavior characteristics, preference characteristics and interaction characteristics to construct a sample user feature set. At the same time, the perceived uniqueness level of each sample user is manually evaluated and scored to form a sample perceived uniqueness score set. Then, using machine learning algorithms such as logistic regression, decision trees, neural networks, etc., with sample user characteristics as input and sample perceived uniqueness scores as target output, the perceived uniqueness evaluator is trained. During the training process, the parameters of the model are continuously adjusted so that its prediction results on the sample data are as close as possible to the actual scores, and the generalization performance of the analyzer, that is, the prediction accuracy on new users, is evaluated by methods such as cross-validation. After training and optimization, a perceived uniqueness evaluator is obtained, which can obtain its perceived uniqueness level according to the user's behavior, preference, and interaction characteristics. Subsequently, for the user to be analyzed currently, first extract its user behavior characteristics, user preference characteristics, and user interaction characteristics to form an input vector isomorphic to the sample user characteristics. Next, the input vector is substituted into the trained perceptual uniqueness evaluator, and after internal calculations of the model, the output result is obtained, which is the perceptual uniqueness score of the user.

[0044] By utilizing machine learning methods, we summarized the rules for perceived uniqueness scoring based on massive user samples and applied them to current users to obtain a quantitative assessment of their perceived uniqueness, thereby improving the efficiency and scalability of perceived uniqueness scoring and providing a basis for subsequent user personalized services.

[0045] Furthermore, the embodiment of the present application also includes:

[0046] Perform a cluster analysis on the users according to the user behavior characteristics, the user preference characteristics and the user interaction characteristics to obtain a regular user group and a high-value user group;

[0047] A secondary clustering analysis is performed on the high-value user group according to the user-perceived uniqueness score to obtain a high-value common user group and a high-value individual user group.

[0048] In a preferred embodiment, first, user behavior characteristics, user preference characteristics and user interaction characteristics are selected as input attributes for clustering. These characteristics can reflect the user's dependence on the platform, consumption ability, interaction intensity, etc., and are important bases for distinguishing user value. Then, clustering algorithms are used to divide users, such as K-means, hierarchical clustering, DBSCAN, etc. In the clustering process, by calculating the similarity between users in the input attributes, users with high similarity are automatically classified into the same cluster. The clustering results divide users into several clusters, and the users in each cluster are similar in behavior, preference, and interaction. Next, the characteristics of each cluster are analyzed and interpreted. Usually, the mean, median and other statistics of each cluster on the clustering attributes are compared, corresponding to the business semantics. For example, clusters with higher consumption frequency and amount are interpreted as "high-value users", clusters with higher interaction frequency but lower consumption amount are interpreted as "highly active low-value users", and clusters with lower consumption and interaction frequency are interpreted as "regular users". According to business goals, several similar clusters are merged to obtain a more macro user group division. For example, clusters such as high consumption and high activity are merged into the "high-value user group", and the remaining clusters are merged into the "regular user group", thus obtaining a coarse-grained binary classification result.

[0049] Subsequently, based on the high-value user group, the high-value user group is further clustered and subdivided to distinguish the strength of the perceived uniqueness of the high-value users, so the user perceived uniqueness score is selected as the input attribute of clustering. Specifically, from the obtained high-value user group, the perceived uniqueness score of each user is extracted, and the high-value users are divided into a high-value individual user group and a high-value common user group through clustering. Among them, the perceived uniqueness score of the high-value individual user group is higher, and the perceived uniqueness score of the high-value common user group is lower.

[0050] Through two-step cluster analysis, users are divided into three groups: regular user group, high-value common user group, and high-value individual user group, laying the foundation for designing different interaction strategies for different groups.

[0051] Furthermore, the embodiment of the present application also includes:

[0052] For the conventional user group, the high-value common user group, and the high-value individual user group, a group user portrait is constructed to obtain a conventional user portrait, a high-value common user portrait, and a high-value individual user portrait;

[0053] The historical user database of the interactive e-commerce platform matches multiple historical users with the conventional user portrait, the high-value common user portrait and the high-value individual user portrait, respectively, to obtain a conventional historical user set, a high-value common historical user set and a high-value individual historical user set;

[0054] The user interaction processes and interaction results of multiple historical users in each historical user set are evaluated, and the conventional interaction strategy template, the high-value common interaction strategy template, and the high-value individual interaction strategy template are generated respectively.

[0055] In a feasible implementation, first, statistical analysis is performed on the user characteristics of each user group to obtain the distribution of each characteristic, such as mean, median, quantile, etc., to find out the common characteristics of the group in terms of behavioral preferences, interaction patterns, etc. Then, combined with business understanding, the typical feature combinations of each group are summarized. After that, these typical feature combinations are abstracted into user portraits. User portraits are a high-level summary of a class of users, describing their basic attributes, behavioral habits, psychological characteristics, etc. The typical feature combinations of the regular user group, high-value common user group, and high-value individual user group obtained in the previous step are summarized as regular user portraits, high-value common user portraits, and high-value individual user portraits, respectively, to form a comprehensive characterization of the three types of users.

[0056] In the historical user database of the e-commerce platform, there are a large number of user records that have completed interactions and purchases. These historical users have similar characteristics to the current users to be marketed, and their past interaction processes and results can provide a reference for current marketing decisions. For each user portrait (conventional user portrait, high-value common user portrait, high-value individual user portrait), search for similar users in the historical user database. For example, similarity is measured by calculating the distance between user feature vectors (such as Euclidean distance, cosine similarity). The closer the historical user is to the current user portrait, the more its feature combination matches the current portrait. For each user portrait, select a number of the most similar historical users to form a conventional historical user set, a high-value common historical user set, and a high-value individual historical user set, which represent the reference samples closest to the current three types of users, and provide valuable prior experience for the marketing decision of the current user.

[0057] Afterwards, for each historical user set, the interaction records between the users and the platform are analyzed one by one. For example, through which channels (such as APP push, SMS, and in-site message) do users mainly interact with the platform, and what are the click-through rates and conversion rates of different channels; to which types of content (such as promotional information, new product recommendations, and personalized retention) do users respond more positively, with higher click-through rates and conversion rates; at what time points (such as holidays and around payday) do users have higher interaction intensity and are more likely to make purchases; what is the user's purchase behavior after the interaction, and how are indicators such as the average order value and repurchase rate performed. By analyzing the interaction data of a large number of historical users, the empirical laws of effective interaction are summarized. These laws can be abstracted into interaction strategy templates to guide the personalized marketing of current users, thereby obtaining conventional interaction strategy templates, high-value common interaction strategy templates, and high-value personalized interaction strategy templates. For example, the analysis of the conventional historical user set found that "pushing promotional information through the APP on weekends can effectively increase the click-through rate and conversion rate", thereby generating a conventional interaction strategy template, that is, for regular users, push promotional information through the APP on weekends.

[0058] Through interaction analysis based on historical users, differentiated interaction strategy templates are generated for different user groups, and specific marketing suggestions for different groups are given to improve the scientificity and effectiveness of marketing decisions.

[0059] Furthermore, the embodiment of the present application also includes:

[0060] Analyzing the group interaction strategy template at the user terminal, extracting user personalized content items of the group interaction strategy template, and forming a user personalized content extraction request;

[0061] Extracting corresponding user personalized data from the user original data set according to the user personalized content extraction request;

[0062] The user personalized data is embedded into the user personalized content item of the group interaction strategy template to form the user personalized interaction strategy.

[0063] In a feasible implementation, when the user terminal receives the interaction strategy template for the group to which the user belongs, first, the group interaction strategy template is parsed to extract the user personalized content items therein, and the user personalized content items need to be filled in according to the specific user information. The user terminal identifies the user personalized content items therein by analyzing the special tags or keywords in the group interaction strategy template, and generates a user personalized content extraction request. The user personalized content extraction request contains the field name to be filled and its corresponding data type, format and other meta information, which is used to guide the subsequent extraction of the corresponding user personalized data from the user's original data set.

[0064] Subsequently, a targeted search is performed in the user's original data set according to the user's personalized content extraction request. Each field in the user's personalized content extraction request is mapped to a specific value in the user's original data set, including different types such as strings, numbers, dates, and image links, to obtain user personalized data, which corresponds one-to-one to the fields in the user's personalized content extraction request. After the required personalized data is extracted from the user's original data set, these user personalized data are filled into the corresponding positions of the group interaction strategy template, replacing the original placeholders, and generating a complete user personalized interaction strategy for the current user. Through the text replacement operation, a general template for the entire user group is transformed into a user personalized interaction strategy for a specific user.

[0065] Furthermore, the embodiment of the present application also includes:

[0066] Executing the user personalized interaction strategy on the user terminal, performing personalized interaction with the user, and synchronously collecting user feedback data on the user personalized interaction strategy during the interaction process;

[0067] The feedback data is analyzed to evaluate the interaction effect of the user personalized interaction strategy, and the user personalized interaction strategy is optimized and adjusted according to the interaction effect.

[0068] In a preferred embodiment, after generating a user personalized interaction strategy, the user personalized interaction strategy is actually executed on the user terminal to conduct real-time, dynamic and personalized interaction with the user. During the interaction process, the user terminal records various user feedback data on the personalized content in real time, including explicit feedback and implicit feedback. Among them, explicit feedback refers to the explicit operations made by users directly on personalized content, such as clicking, browsing, collecting, commenting, purchasing, etc., which intuitively reflect the user's interest and acceptance of the content; implicit feedback refers to the user's lack of direct operation, but his behavior pattern can indirectly reflect his attitude towards the content, such as browsing time, browsing depth, page bounce rate, etc., which can infer the user's actual feelings about the content. In addition to behavioral feedback, subjective feedback from users can also be collected, such as inviting users to rate personalized recommendations, provide improvement suggestions, etc., to reflect the user's true thoughts.

[0069] Then, the user feedback data of the user personalized interaction strategy is summarized and counted, and the effect index is calculated to obtain the interaction effect of the user personalized interaction strategy. For example, for explicit feedback, the click-through rate, conversion rate, purchase amount, etc. are counted; for implicit feedback, the average browsing time, exit rate, etc. are counted; for subjective feedback, the average score, positive evaluation ratio, etc. are counted. After that, the current user personalized interaction strategy is compared with the historical user personalized interaction strategy, and the current interaction effect and the historical interaction effect are analyzed to judge the pros and cons of the current user personalized interaction strategy, identify the problems of the current user personalized interaction strategy, and optimize the user personalized interaction strategy in a targeted manner. For the user personalized interaction strategy with poor interaction effect, it is directly eliminated or significantly adjusted; for the user personalized interaction strategy with room for optimization, it is optimized and fine-tuned in terms of content quality, form presentation, push timing, etc., for example, replace the content material that is closer to the user's interests, simplify the text and pictures presented in the content, and adjust the time and frequency of content push. Through continuous strategy optimization and iteration, the overall effect of personalized strategy is continuously improved, and the level of personalized marketing intelligence is improved.

[0070] In summary, the marketing decision-making method for an e-commerce platform provided in the embodiments of the present application has the following technical effects:

[0071] Send a data collection instruction to the user terminal, collect the user's browsing behavior data and transaction data at the user terminal, form a user original data set, obtain comprehensive and real user behavior data, and lay a data foundation for subsequent steps; receive the user original data set sent by the user terminal, perform data cleaning and feature extraction, and construct a user feature vector containing the user's perceived uniqueness score to provide a basis for user segmentation. Perform cluster analysis on the user feature vector to divide the users into a regular user group, a high-value common user group, and a high-value individual user group, wherein the user's perceived uniqueness score of the high-value individual user group is higher than that of the high-value common user group, to achieve user segmentation, distinguish high-value user groups according to the user's perceived uniqueness score, and lay a foundation for differentiated marketing. Generate corresponding group interaction strategy templates for the regular user group, the high-value common user group, and the high-value individual user group, respectively, which are the regular interaction strategy template, the high-value common interaction strategy template, and the high-value individual interaction strategy template. According to the characteristics of different user groups, design differentiated interaction strategy templates to reflect the personalization of marketing decisions. The corresponding group interaction strategy template is sent to the user terminal, and the group interaction strategy template and the user original data set are combined at the user terminal to generate a user personalized interaction strategy. The interaction strategy template at the group level is combined with the user's individual data to generate a personalized interaction strategy for each user, thereby improving the accuracy and effectiveness of marketing decisions and meeting the differentiated needs of different users.

[0072] Embodiment 2 is based on the same inventive concept as the marketing decision-making method for an e-commerce platform in the above embodiment. Figure 2 As shown, the embodiment of the present application provides a marketing decision system for an e-commerce platform, the system comprising:

[0073] The data collection module 11 is used to send a data collection instruction to the user terminal, collect the user's browsing behavior data and transaction data at the user terminal, and form a user original data set;

[0074] The data processing module 12 is used to receive the original user data set sent by the user terminal, perform data cleaning and feature extraction, and construct a user feature vector containing a user perceived uniqueness score;

[0075] A user classification module 13 is used to perform cluster analysis on the user feature vectors to classify the users into a regular user group, a high-value commonality user group and a high-value individuality user group, wherein the user perceived uniqueness score of the high-value individuality user group is higher than that of the high-value commonality user group;

[0076] A strategy template generating module 14 is used to generate corresponding group interaction strategy templates for the conventional user group, the high-value commonality user group, and the high-value individuality user group, which are respectively a conventional interaction strategy template, a high-value commonality interaction strategy template, and a high-value individuality interaction strategy template;

[0077] The user strategy generation module 15 is used to send the corresponding group interaction strategy template to the user terminal, and generate a user personalized interaction strategy by combining the group interaction strategy template and the user original data set at the user terminal.

[0078] Furthermore, the data processing module 12 includes the following execution steps:

[0079] After data cleaning and feature extraction of the original user data set, user behavior features, user preference features and user interaction features are obtained;

[0080] Obtaining a user perceived uniqueness score according to the user behavior characteristics, the user preference characteristics, and the user interaction characteristics;

[0081] The user behavior characteristics, the user preference characteristics, the user interaction characteristics and the user perceived uniqueness score are aggregated to generate the user feature vector.

[0082] Furthermore, the data processing module 12 further includes the following execution steps:

[0083] A historical user database of the interactive e-commerce platform, based on which a sample user feature set and a sample perceived uniqueness score set are constructed, wherein the sample user features include sample user behavior features, sample user preference features, and sample user interaction features;

[0084] Constructing a perceptual uniqueness evaluator based on the sample user feature set and the sample perceptual uniqueness score set;

[0085] The user behavior characteristics, the user preference characteristics and the user interaction characteristics are input into a perceived uniqueness evaluator to obtain a user perceived uniqueness score.

[0086] Furthermore, the user segmentation module 13 includes the following execution steps:

[0087] Perform a cluster analysis on the users according to the user behavior characteristics, the user preference characteristics and the user interaction characteristics to obtain a regular user group and a high-value user group;

[0088] A secondary clustering analysis is performed on the high-value user group according to the user-perceived uniqueness score to obtain a high-value common user group and a high-value individual user group.

[0089] Furthermore, the policy template generation module 14 includes the following execution steps:

[0090] For the conventional user group, the high-value common user group, and the high-value individual user group, a group user portrait is constructed to obtain a conventional user portrait, a high-value common user portrait, and a high-value individual user portrait;

[0091] The historical user database of the interactive e-commerce platform matches multiple historical users with the conventional user portrait, the high-value common user portrait and the high-value individual user portrait, respectively, to obtain a conventional historical user set, a high-value common historical user set and a high-value individual historical user set;

[0092] The user interaction processes and interaction results of multiple historical users in each historical user set are evaluated, and the conventional interaction strategy template, the high-value common interaction strategy template, and the high-value individual interaction strategy template are generated respectively.

[0093] Furthermore, the user policy generation module 15 includes the following execution steps:

[0094] Analyzing the group interaction strategy template at the user terminal, extracting user personalized content items of the group interaction strategy template, and forming a user personalized content extraction request;

[0095] Extracting corresponding user personalized data from the user original data set according to the user personalized content extraction request;

[0096] The user personalized data is embedded into the user personalized content item of the group interaction strategy template to form the user personalized interaction strategy.

[0097] Furthermore, the embodiment of the present application also includes a user policy optimization module, which includes the following execution steps:

[0098] Executing the user personalized interaction strategy on the user terminal, performing personalized interaction with the user, and synchronously collecting user feedback data on the user personalized interaction strategy during the interaction process;

[0099] The feedback data is analyzed to evaluate the interaction effect of the user personalized interaction strategy, and the user personalized interaction strategy is optimized and adjusted according to the interaction effect.

[0100] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0101] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these modifications and variations.

Claims

1. A marketing decision-making method for an e-commerce platform, characterized in that: The method comprises: Sending data collection instructions to the user terminal to collect the user's browsing behavior data and transaction data at the user terminal to form the user's original data set; Receiving the user original data set sent by the user terminal, performing data cleaning and feature extraction, and constructing a user feature vector including a user perceived uniqueness score; Performing cluster analysis on the user feature vectors to divide the users into a regular user group, a high-value commonality user group, and a high-value individuality user group, wherein the user perceived uniqueness score of the high-value individuality user group is higher than that of the high-value commonality user group; Generate corresponding group interaction strategy templates for the conventional user group, the high-value commonality user group, and the high-value individuality user group, which are respectively a conventional interaction strategy template, a high-value commonality interaction strategy template, and a high-value individuality interaction strategy template; Sending a corresponding group interaction strategy template to a user terminal, and combining the group interaction strategy template with the user original data set at the user terminal to generate a user personalized interaction strategy; Construct a user feature vector containing the user's unique perceived ratings, including: After data cleaning and feature extraction of the original user data set, user behavior features, user preference features and user interaction features are obtained; Obtaining a user perceived uniqueness score according to the user behavior characteristics, the user preference characteristics, and the user interaction characteristics; Aggregating the user behavior features, the user preference features, the user interaction features, and the user perceived uniqueness score to generate the user feature vector; Obtaining a user perceived uniqueness score according to the user behavior characteristics, the user preference characteristics, and the user interaction characteristics, including: A historical user database of the interactive e-commerce platform, based on which a sample user feature set and a sample perceived uniqueness score set are constructed, wherein the sample user features include sample user behavior features, sample user preference features, and sample user interaction features; Constructing a perceptual uniqueness evaluator based on the sample user feature set and the sample perceptual uniqueness score set; Inputting the user behavior feature, the user preference feature and the user interaction feature into a perceived uniqueness evaluator to obtain a user perceived uniqueness score; Performing cluster analysis on the user feature vectors to divide the users into a regular user group, a high-value common user group, and a high-value individual user group, including: Perform a cluster analysis on the users according to the user behavior characteristics, the user preference characteristics and the user interaction characteristics to obtain a regular user group and a high-value user group; A secondary clustering analysis is performed on the high-value user group according to the user-perceived uniqueness score to obtain a high-value common user group and a high-value individual user group.

2. The marketing decision-making method for an e-commerce platform according to claim 1, characterized in that: For the conventional user group, the high-value commonality user group, and the high-value individuality user group, corresponding group interaction strategy templates are generated, which are respectively a conventional interaction strategy template, a high-value commonality interaction strategy template, and a high-value individuality interaction strategy template, including: For the conventional user group, the high-value common user group, and the high-value individual user group, a group user portrait is constructed to obtain a conventional user portrait, a high-value common user portrait, and a high-value individual user portrait; The historical user database of the interactive e-commerce platform matches multiple historical users with the conventional user portrait, the high-value common user portrait and the high-value individual user portrait, respectively, to obtain a conventional historical user set, a high-value common historical user set and a high-value individual historical user set; The user interaction processes and interaction results of multiple historical users in each historical user set are evaluated, and the conventional interaction strategy template, the high-value common interaction strategy template, and the high-value individual interaction strategy template are generated respectively.

3. The marketing decision-making method for an e-commerce platform according to claim 1, characterized in that: Sending a corresponding group interaction strategy template to a user terminal, and combining the group interaction strategy template with the user original data set at the user terminal to generate a user personalized interaction strategy, including: Analyzing the group interaction strategy template at the user terminal, extracting user personalized content items of the group interaction strategy template, and forming a user personalized content extraction request; Extracting corresponding user personalized data from the user original data set according to the user personalized content extraction request; The user personalized data is embedded into the user personalized content item of the group interaction strategy template to form the user personalized interaction strategy.

4. The marketing decision-making method for an e-commerce platform according to claim 1, characterized in that: The method further comprises: Executing the user personalized interaction strategy on the user terminal, performing personalized interaction with the user, and synchronously collecting user feedback data on the user personalized interaction strategy during the interaction process; The feedback data is analyzed to evaluate the interaction effect of the user personalized interaction strategy, and the user personalized interaction strategy is optimized and adjusted according to the interaction effect.

5. A marketing decision system for an e-commerce platform, characterized in that: A system for implementing the marketing decision-making method for an e-commerce platform according to any one of claims 1 to 4, comprising: A data collection module, which is used to send a data collection instruction to a user terminal, collect the user's browsing behavior data and transaction data at the user terminal, and form a user original data set; A data processing module, the data processing module is used to receive the user original data set sent by the user terminal, perform data cleaning and feature extraction, and construct a user feature vector containing the user perceived uniqueness score; A user segmentation module, the user segmentation module is used to perform cluster analysis on the user feature vectors, and to segment the users into a regular user group, a high-value commonality user group, and a high-value individuality user group, wherein the user perceived uniqueness score of the high-value individuality user group is higher than that of the high-value commonality user group; A strategy template generation module, wherein the strategy template generation module is used to generate corresponding group interaction strategy templates for the conventional user group, the high-value commonality user group, and the high-value individuality user group, which are respectively a conventional interaction strategy template, a high-value commonality interaction strategy template, and a high-value individuality interaction strategy template; A user strategy generation module is used to send a corresponding group interaction strategy template to a user terminal, and generate a user personalized interaction strategy by combining the group interaction strategy template and the user original data set at the user terminal.

Images