Background operation management system based on big data

Abstract

The invention relates to the technical field of data management, in particular to a background operation management system based on big data, and the system comprises a data analysis unit which is used for obtaining first interaction data of a customer, and determining a first interest preference feature of the customer based on the first interaction data; wherein the first interaction data comprises a browsing record, a search record and a purchase record of the customer on the marketing platform; the group classification unit is used for carrying out clustering analysis on the first interest preference characteristics and determining interest group categories of the clients; wherein the interest group categories are divided according to the similarity degree of customer interests. By analyzing the interaction data of the clients, the system can accurately identify the interests and preferences of the clients and classify the clients into different interest groups through clustering analysis, and the personalized recommendation based on the data can make marketing resources and activities more targeted, so that the participation degree and satisfaction degree of the clients are improved.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a back-end operation management system based on big data. Background Technology

[0002] In today's fiercely competitive market, businesses need to respond to changes in market demand in a very short time. Traditional operating models often rely on historical experience and limited real-time data, making it difficult to make quick decisions. Big data systems, on the other hand, can provide businesses with precise decision support by collecting and analyzing market, customer, and competitor data in real time. This helps businesses flexibly adjust their strategies, seize market opportunities, and reduce the risk of decision-making errors.

[0003] Currently, traditional systems often rely on simple customer data, such as age, gender, and region, for customer segmentation without in-depth analysis of customer interests and preferences. This can result in recommendation systems pushing content that doesn't match actual customer needs, failing to achieve personalized recommendations, reducing the effectiveness of marketing campaigns and customer engagement. Furthermore, they typically lack the ability to adjust marketing strategies based on real-time customer feedback. When customers don't respond to certain marketing content, the system cannot update their preference data promptly, leading to wasted resources. When customer behavior changes, traditional systems often cannot react quickly, and the reliance on static data makes marketing strategies lag behind.

[0004] Furthermore, traditional systems often push marketing content based on fixed customer tags without considering customer behavior data and changes in interests. This makes it difficult for traditional systems to efficiently match marketing resources, resulting in low relevance between pushed content and customer interests, and poor marketing campaign performance. Moreover, they usually rely on a single feature to segment customer groups, lacking in-depth analysis of customer interests and preferences, and failing to refine the specific interest levels of customers for different products, brands, or price ranges. This makes it impossible for the system to make personalized and accurate recommendations when formulating marketing strategies. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: a back-end operation management system based on big data, comprising: The data analysis unit is used to acquire the customer's first interaction data and determine the customer's first interest and preference characteristics based on the first interaction data; wherein, the first interaction data includes the customer's browsing history, search history and purchase history on the marketing platform; A group classification unit is used to perform cluster analysis on the first interest preference features to determine the customer's interest group category; wherein the interest group category is divided according to the similarity of customer interests; The resource filtering unit is used to filter out a first set of marketing resources from the marketing resource library based on the interest group category; wherein, the first set of marketing resources includes marketing content, marketing activity plans and marketing promotion information related to the interest group category; The resource matching unit is used to evaluate the matching degree between the first marketing resource set and the interest group category, and obtain the matching degree evaluation result. An operations management unit is used to push the first set of marketing resources to customers in the corresponding interest group category in response to the matching degree evaluation result indicating a high matching degree.

[0006] Preferably, determining the customer's first interest preference characteristics based on the first interaction data includes: Multiple basic features are extracted from the first interaction data, and correlation analysis is performed on the multiple basic features to obtain a combination of related features; wherein, the basic features include the browsed product categories, search keywords, and the price range of the purchased products; Based on the aforementioned combination of related features, cluster analysis is performed on customers to obtain multiple interest groups; each interest group includes customers with similar interest features. For each interest group, analyze its corresponding combination of related features to determine the primary interest preference feature for each interest group, and obtain the primary interest preference feature for each customer.

[0007] Preferably, correlation analysis is performed on the multiple basic features to obtain a combination of related features, including: Analyze the semantic relationship between browsed product categories and search keywords to determine semantically consistent combinations of browsed product categories and search keywords; Analyze the price matching relationship between the price range of purchased goods and the product categories browsed, and determine the price matching combination of the purchased goods price range and the product categories browsed; By fusing combinations with semantic consistency with combinations with price matching, we obtain combinations of related features; Analyze the browsing order of different product categories in the browsing history to determine the customer's preferred order of browsing products; By associating the order in which products are viewed with the product categories, a combination of associated features, including the browsing order, is obtained.

[0008] Preferably, for each interest group, the corresponding combination of related features is analyzed to determine the first interest preference feature corresponding to each interest group, thus obtaining the first interest preference feature of each customer, including: For each interest group, the frequency of occurrence of each basic feature in its associated feature combination is counted, and the basic feature with the highest frequency is determined as the main interest feature of the interest group. Based on the main interest features and other features in the combination of related features, the first interest preference feature corresponding to the interest group is determined; Based on the customer's interest group, the first interest preference feature corresponding to the interest group is assigned to the customer, thus obtaining the first interest preference feature of each customer; For each interest group, analyze the correlation strength between different features in its associated feature combination, and determine the hierarchical structure of interest preferences based on the correlation strength; Based on the hierarchical structure of interest preferences, the first interest preference features corresponding to the interest groups are refined.

[0009] Preferably, the first interest preference feature includes multiple interest dimensions, the interest group category includes multiple interest subgroups, and the interest dimensions include product type interest, brand interest, and price range interest. Cluster analysis is performed on the first interest preference feature to determine the customer's interest group categories, including: Customers with similar product type interests, similar brand interests, and similar price range interests are identified as a group across multiple interest dimensions. Interest subgroups are determined based on the degree of similarity among each group of customers across various interest dimensions.

[0010] Preferably, the method further includes: In response to detecting customer feedback data on a first set of marketing resources pushed to the system, a second set of customer interest preferences is obtained based on the feedback data; wherein the feedback data includes the customer's click-through rate, dwell time, and enthusiasm for participating in marketing activities.

[0011] Preferably, a matching degree assessment is performed on the first marketing resource set and the interest group category to obtain a matching degree assessment result, including: Based on the historical feedback data of customers in the interest group category to marketing resources, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of the interest group category, the matching degree is compared with the preset matching degree threshold to determine the matching degree evaluation result. Specifically, if the proportion of positive feedback in the historical feedback data of customers in the interest group category to marketing resources is greater than a preset first matching degree threshold, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of the interest group category is greater than a preset second matching degree threshold, then the matching degree assessment result is judged to be a high matching degree.

[0012] Preferably, the method further includes: In response to the absence of detected customer feedback data on the first set of marketing resources pushed to the system, supplementary customer interaction data is obtained, wherein the supplementary interaction data includes customer behavior data on other relevant platforms; Based on the supplementary interaction data, the customer's supplementary interest preference characteristics are determined, and the interest group category is updated based on the supplementary interest preference characteristics to obtain the updated interest group category; Based on the updated interest group categories, a second set of marketing resources is selected from the marketing resource library; A new matching assessment is performed on the second set of marketing resources and the updated interest group categories; In response to determining that the result of the new matching assessment is a high matching degree, the second set of marketing resources is pushed to customers in the corresponding updated interest group category.

[0013] Preferably, the matching degree evaluation of the first marketing resource set and the interest group category to obtain the matching degree evaluation result further includes: If, within the interest group category, the proportion of positive feedback from customers in each interest subgroup regarding marketing resources in historical feedback data is greater than a preset first matching threshold, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of each interest subgroup is greater than a preset second matching threshold, then the matching degree assessment result is determined to be a high matching degree.

[0014] Preferably, the matching degree evaluation of the first marketing resource set and the interest group category to obtain the matching degree evaluation result further includes: If, within an interest group category, the proportion of positive feedback on marketing resources from customers within a subgroup is less than the first matching threshold, then the matching assessment result is determined to be a low matching degree, and marketing resources are re-screened for that subgroup; or... If there is a marketing resource in the first marketing resource set whose matching degree with the interest points of the interest subgroup is less than the second matching degree threshold, then the matching degree assessment result is determined to be low matching degree, and the marketing resource is adjusted or replaced.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention analyzes customer interaction data, enabling the system to accurately identify customer interests and preferences. Through cluster analysis, customers are categorized into different interest groups. This data-driven personalized recommendation makes marketing resources and activities more targeted, thereby increasing customer engagement and satisfaction. Furthermore, by assessing the match between the characteristics of interest groups and marketing resources, the system ensures that the pushed marketing content, activities, and promotional information highly align with customers' actual interests, improving the effectiveness and conversion rate of marketing activities. Moreover, by using historical feedback data and matching thresholds for matching evaluation, the system can more accurately push highly matched marketing resources to customers, increasing customer response rates to marketing content and ultimately improving sales conversion rates. This invention dynamically adjusts marketing resources based on customer feedback data. If a customer does not respond to the pushed marketing resources, the system can also obtain supplementary interaction data, update customer interests and preferences in real time, and re-select marketing resources. This feedback mechanism can effectively improve customer engagement and avoid resource waste. Moreover, by performing correlation analysis on multiple basic features, the system can uncover deeper characteristics of customer interests, including multiple dimensions such as product type, brand, and price range. Based on cluster analysis of these dimensions, the system can not only identify the general interest groups of customers, but also refine the interest levels of each customer, further refining marketing strategies. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall system architecture in one embodiment of the present invention.

[0017] In the diagram: 1. Data analysis unit; 2. Group classification unit; 3. Resource screening unit; 4. Resource matching unit; 5. Operation management unit. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 This invention provides a technical solution: a back-end operation management system based on big data, comprising: Data analysis unit 1 is used to acquire the customer's first interaction data and determine the customer's first interest and preference characteristics based on the first interaction data; wherein, the first interaction data includes the customer's browsing history, search history and purchase history on the marketing platform; Group classification unit 2 is used to perform cluster analysis on the first interest preference feature to determine the customer's interest group category; wherein, the interest group category is divided according to the similarity of customer interests; Resource filtering unit 3 is used to filter out a first set of marketing resources from the marketing resource library based on interest group categories; wherein, the first set of marketing resources includes marketing content, marketing activity plans and marketing promotion information related to interest group categories; Resource matching unit 4 is used to evaluate the matching degree between the first marketing resource set and the interest group category, and obtain the matching degree evaluation result; Operations management unit 5 is used to push the first set of marketing resources to customers in the corresponding interest group category in response to the matching evaluation result indicating a high matching degree.

[0020] It's important to note that acquiring customers' initial interaction data and using this data to determine their interests and preferences is crucial. This initial interaction data includes customers' browsing history, search history, and purchase history on the marketing platform. For example, suppose you're an operations staff member of an online retail platform. In this system, when a customer logs in for the first time, the system automatically records the customer's browsing behavior, such as viewing shoes from a particular brand, their search history (e.g., searching for "summer sneakers"), and their purchase history (e.g., buying a pair of blue sneakers). This data constitutes the customer's "initial interaction data." Cluster analysis is performed on collected customer interest and preference characteristics to divide customers into different interest groups, usually based on the similarity of customer interests. Cluster analysis can help businesses identify different customer groups and thus develop personalized marketing strategies. For example, suppose many customers have purchased sneakers, yoga mats, and sportswear on the platform; these customers' interest group can be defined as the "sports enthusiasts" group. Customers who frequently purchase fashion clothing and cosmetics would be classified as the "fashion enthusiasts" group. In this way, the system divides customers into different groups based on their interests and preferences. Based on the customer's interest group category, the system filters out a set of marketing resources related to that interest group from the marketing resource library. This set includes marketing content, activity plans, and promotional information relevant to the customer's interest group. For example, for the "sports enthusiast" group, the system will filter out relevant sports product marketing content from the marketing resource library, such as introductions to new running shoes, and provide corresponding promotional activities, such as discounts from sports brands, and coupons applicable to this group, such as a 50 RMB discount on sports equipment purchases over 200 RMB. Similarly, for the "fashion enthusiast" group, the system will filter out the latest fashion product information and related fashion discounts. The system evaluates the matching degree between the selected marketing resource sets and customer group categories, resulting in a matching degree assessment for each resource set and the target customer group. A higher matching degree indicates that the marketing resource is more suitable for the group. For example, suppose a series of sports products and promotional activities are selected for the "sports enthusiast" group. The system will assess whether these products and activities meet the group's needs. For instance, if a customer has viewed running shoe-related pages multiple times without purchasing any products, the system will consider the running shoe-related marketing activities to be highly matched with the customer's interests, resulting in a high matching degree. For customers with low interest, such as those who have only viewed the running shoe page once, the matching degree will be lower. Based on the matching assessment results, marketing resources are pushed to customer groups with high matching scores. These resources typically include customized advertisements, promotional information, coupons, etc. For example, if the matching assessment result of a certain "sports enthusiast" group is high, the operations management unit will automatically push marketing resources related to this group (such as promotional activities and coupons for sports shoes) to these customers. Specifically, customers will see information about running shoe discounts on the platform or receive relevant promotional information through email, SMS, and other channels.

[0021] In an optional embodiment, determining a customer's first interest preference feature based on first interaction data includes: Multiple basic features are extracted from the first interaction data, and correlation analysis is performed on these basic features to obtain a combination of related features. Among them, the basic features include the browsing product categories, search keywords, and the price range of the purchased products. Cluster analysis of customers is performed based on combinations of related features to obtain multiple interest groups; each interest group includes customers with similar interest features. For each interest group, analyze its corresponding combination of related features to determine the primary interest preference feature for each interest group, and obtain the primary interest preference feature for each customer.

[0022] It's important to note that several fundamental features need to be extracted from the customer's initial interaction data. This data includes the customer's browsing, searching, and purchasing behaviors on the platform. These fundamental features typically include: Product categories browsed: What types of products the customer viewed, such as electronics, clothing, and cosmetics; Search keywords: The keywords the customer entered in the search box, such as "summer athletic shoes" or "high-end laptops"; Purchase price range: The price range of the products the customer purchased, such as "100-200 yuan" or "2000-3000 yuan". For example, suppose a customer browsed several athletic shoes (belonging to the "sports equipment" category) on the platform, searched for "running shoes" and "fitness equipment," and purchased a pair of running shoes priced at 150 yuan. For this customer, the fundamental features might include: Product category browsed: Sports equipment; Search keywords: running shoes, fitness equipment; Purchase price range: 100-200 yuan. Multiple basic features are analyzed for correlation to identify which features are strongly correlated. For example, certain product categories and search keywords may frequently appear together, or purchasing behavior within certain price ranges may be related to specific product categories. For instance, through analysis, the system may find a strong correlation between browsing behavior in the "sports equipment" category and the search keywords "running shoes" and "fitness equipment." Furthermore, customers who purchase items in the "100-200 yuan" price range often also buy running shoes. Therefore, a possible combination of related features might be: browsing product category = sports equipment, search keywords = running shoes, and purchase price range = 100-200 yuan. Based on these associated features, the system performs cluster analysis on customers. Cluster analysis groups customers with similar interests into a single group. Each group of customers shares similar purchasing behaviors and interest characteristics. For example, assuming there are a large number of customers on the platform, the system divides them into several interest groups through cluster analysis. For instance, Group A: all customers who browsed the "Sports Goods" category and searched for "Running Shoes," and most of these customers' purchase prices are between 100-200 yuan; Group B: customers who browsed the "Fashion Goods" category and searched for "Trendy Bags," and their purchase prices are between 500-1000 yuan. Each group represents a type of customer with similar interests. Based on the cluster analysis results, the system analyzes the combination of associated characteristics for each interest group and determines the corresponding primary interest preference characteristics for each group. These characteristics will help the system provide customized marketing resources for that group. For example: Group A (sports enthusiasts): The primary interest preference characteristics of this group may be "running shoes" and "fitness equipment". According to the analysis, these customers usually buy sports products priced between 100-200 yuan, so they may be interested in discount promotions, new sports equipment, etc. Group B (fashion enthusiasts): The primary interest preference characteristics of this group may be "trendy bags" and "fashionable clothing", and they tend to buy products priced between 500-1000 yuan. Therefore, they may be interested in new fashion products, limited-time discounts, etc. Ultimately, through these cluster analyses and feature extractions, each customer is assigned to a specific interest group, and their primary interest preference characteristics are derived based on the characteristics of that group. Each customer's preference characteristics will help the platform push relevant marketing content and product recommendations to them. For example, if a customer's behavior matches the characteristics of group A, such as browsing sports equipment, searching for running shoes, and purchasing running shoes priced at 150 yuan, then that customer's primary interest preference characteristics are "running shoes" and "fitness equipment," which will influence the marketing information and recommended products pushed to them by the platform.

[0023] In an optional embodiment, association analysis is performed on multiple basic features to obtain a combination of associated features, including: Analyze the semantic relationship between browsed product categories and search keywords to determine semantically consistent combinations of browsed product categories and search keywords; Analyze the price matching relationship between the price range of purchased goods and the product categories browsed, and determine the price matching combination of the purchased goods price range and the product categories browsed; By fusing combinations with semantic consistency with combinations with price matching, we obtain combinations of related features; Analyze the browsing order of different product categories in the browsing history to determine the customer's preferred order of browsing products; By associating the order in which products are viewed with the product categories, a combination of associated features, including the browsing order, is obtained.

[0024] It's important to note that, firstly, it's necessary to analyze whether there's a semantic relationship between the product categories a customer browses and the keywords they search for. Semantic consistency refers to the semantic relevance between certain product categories and search keywords; that is, the customer's search keywords match the product categories they browse, forming a meaningful combination. For example: a customer browses the product category "sports shoes"; the customer searches for the keyword "running shoes." In this example, there's a strong semantic relationship between "sports shoes" and "running shoes" because running shoes are a type of sports shoe. Such a combination of browsed product categories and search keywords demonstrates semantic consistency. For example: a customer browses the product category "electronic products"; the customer searches for the keyword "headphones." Here, "electronic products" and "headphones" also have semantic consistency because headphones are a type of electronic product. Through this analysis, product categories relevant to the customer's search keywords can be recommended. Next, it's necessary to analyze the relationship between customers' purchase price ranges and the product categories they browse, determining which price ranges match which product categories. This is to identify the match between customers' spending power and areas of interest. For example: a customer browses the product category "smartphones"; the customer's purchase price range is 3000-5000 yuan. In this example, the customer browsed the smartphone category and purchased a phone priced roughly within the 3000-5000 yuan range. This indicates a strong match between the price range and the browsed product categories. For example: a customer browses the product category "luxury handbags"; the customer's purchase price range is 5000-10000 yuan. Here, the browsing behavior for luxury handbags matches the 5000-10000 yuan price range, indicating that the customer prefers to buy high-priced goods. Through such analysis, price-matching combinations can be derived, helping the platform to make accurate product recommendations. By fusing semantic consistency combinations and price-matching combinations, a more refined and precise set of related features is obtained. This allows the system to not only know what types of products customers are searching for, but also what price range they are willing to pay for those products. For example, suppose a customer browses the product category "sports shoes"; searches for the keyword "running shoes"; and has a purchase price range of 100-200 yuan. Through association analysis, the system will fuse these features into a single related feature combination: Browsing category = sports shoes; Search keyword = running shoes; Purchase price range = 100-200 yuan. This fusion helps the platform accurately understand customers' interests and budgets, thereby recommending more products that match their preferences. The order in which customers browse products can also provide valuable information. By analyzing the order in which customers browse products, we can understand their browsing preferences. The priority order of browsing certain product categories may indicate that customers have a stronger interest in these categories. For example, suppose a customer's browsing history is as follows: browsing product category: athletic shoes; browsing product category: fitness equipment; browsing product category: sportswear. From this browsing order, we can infer the customer's interest order: athletic shoes > fitness equipment > sportswear. This order shows that the customer is more inclined to focus on athletic shoes, while sportswear and fitness equipment are secondary interests. Finally, the browsing order is combined with the product category to obtain a combination of related features with browsing order. This combination can reveal the dynamic preferences and interests of customers when browsing different product categories. For example, suppose a customer's browsing history is: browsing product category: athletic shoes; browsing product category: sportswear; browsing product category: fitness equipment. The following combination of related features can be obtained: browsing order feature combination: athletic shoes > sportswear > fitness equipment. Based on this analysis, the platform can infer that the customer has the strongest interest in the "athletic shoes" category, and thus prioritize recommending athletic shoe-related products to them, such as running shoes and new athletic shoes.

[0025] In an optional embodiment, for each interest group, its corresponding combination of related features is analyzed to determine the first interest preference feature corresponding to each interest group, thereby obtaining the first interest preference feature of each customer, including: For each interest group, the frequency of occurrence of each basic feature in its associated feature combination is counted, and the basic feature with the highest frequency is determined as the main interest feature of the interest group. Based on the main interest features and other features in the combination of related features, the first interest preference feature corresponding to the interest group is determined. Based on the customer's interest group, the first interest preference feature corresponding to the interest group is assigned to the customer, thus obtaining the first interest preference feature of each customer; For each interest group, analyze the correlation strength between different features in its associated feature combination, and determine the hierarchical structure of interest preferences based on the correlation strength; Based on the hierarchical structure of interest preferences, the primary interest preference features corresponding to interest groups are refined.

[0026] It's important to note that, firstly, for each interest group, it's necessary to statistically analyze the frequency of each basic feature (such as product category, price range, search keywords, etc.) in the associated feature combinations. The basic feature with the highest frequency can be considered the main interest feature of that interest group. For example, suppose we analyze the behavior of a group of sports enthusiasts. Here are some associated feature combinations for this group: Combination 1: Browsing product category = sports shoes, purchase price range = 300-500 yuan, search keyword = running shoes; Combination 2: Browsing product category = fitness equipment, purchase price range = 500-1000 yuan, search keyword = dumbbells; Combination 3: Browsing product category = sportswear, purchase price range = 200-400 yuan, search keyword = sports T-shirts. In the analysis of this group, sports shoes appear most frequently as the browsing product category, therefore, sports shoes can be identified as the main interest feature of this group. After identifying the primary interest characteristics, combining them with other characteristics, such as price range and search keywords, can further determine the group's primary interest preference characteristics. For example, continuing the previous example, we can infer that the primary interest preference characteristic of this sports enthusiast group is athletic shoes (running shoes), and the price range is likely concentrated in the 300-500 yuan range, with search keywords focused on running shoes. Therefore, the group's primary interest preference characteristics are: Primary interest preference characteristics: browsing product category = athletic shoes; price range = 300-500 yuan; search keywords = running shoes. Based on this information, we can infer that the core interest of this group is the product type of running shoes. Based on the customer's interest group, a primary interest preference characteristic is assigned to each customer, ensuring that each customer's interest characteristics are accurately captured and defined. For example, if a customer is highly similar to the sports enthusiast group in multiple behavioral data points, such as having browsed a large number of sports shoes, searched for running shoes, and purchased sports shoes priced between 300-500 yuan, then this customer will be classified as a sports enthusiast and assigned a primary interest preference characteristic for this group: Customer's primary interest preference characteristics: Browsed product category = sports shoes; Price range = 300-500 yuan; Search keyword = running shoes; In this way, a clear primary interest preference characteristic is determined for each customer. Within each interest group, different characteristics may exhibit varying degrees of correlation. For instance, some characteristics may play a dominant role in group behavior, while others may have a secondary role. Analyzing the correlation strength between these characteristics is necessary to determine the hierarchical structure of interest preferences for each group. For example, assuming we continue analyzing the sports enthusiast group: correlation analysis might show a strong correlation between browsing product categories = sports shoes and searching keywords = running shoes, indicating that sports shoes and running shoes are the core interests of this group. Price range may have a weaker correlation with interests and serve only as a reference for purchasing behavior. Based on this analysis, the hierarchical structure might be as follows: First level (core interest): browsing product categories = sports shoes; searching keywords = running shoes; Second level (secondary characteristics): price range = 300-500 yuan; Finally, based on the hierarchical structure of interests and preferences described above, the primary interest and preference characteristics of each group can be further refined. This helps to make more precise divisions of customer interests and provide more accurate personalized recommendations. For example, by analyzing the sports enthusiast group, more precise primary interest and preference characteristics may be derived: Primary interest and preference characteristics: First level: Browsing product category = sports shoes; Search keywords = running shoes; Second level: Purchase price range = 300-500 yuan. In this way, suitable products can be recommended to the customer more accurately. For example, the system will prioritize recommending running shoes to the customer, and then recommend sports shoes with a price range of 300-500 yuan.

[0027] In an optional embodiment, the first interest preference feature includes multiple interest dimensions, the interest group category includes multiple interest subgroups, and the interest dimensions include product type interest, brand interest, and price range interest. Cluster analysis is performed on the primary interest preference characteristics to determine the customer's interest group categories, including: Customers with similar product type interests, similar brand interests, and similar price range interests are identified as a group across multiple interest dimensions. Interest subgroups are determined based on the degree of similarity among each group of customers across various interest dimensions.

[0028] It's important to note that the first interest preference feature doesn't just include a single interest characteristic, but rather encompasses multiple interest dimensions. These dimensions help to more comprehensively understand customer interests. Specific interest dimensions include: Product type interest: customer preferences for different product categories; for example, sneakers, mobile phones, televisions, etc.; Brand interest: customer preferences for specific brands; for example, Nike, Apple, Samsung, etc.; Price range interest: customer preferences for different price ranges; for example, below 200 yuan, 200-500 yuan, 500-1000 yuan, etc. Each customer exhibits different behaviors across these interest dimensions, and the combination of these dimensions helps to accurately depict customer interests. When performing cluster analysis, clusters are determined based on the similarity of customers across each interest dimension. Specifically, customers with similar product type interests, similar brand interests, and similar price range interests are clustered into the same group. For example, suppose we analyze the data of the following three customers, focusing on their interests in product type, brand, and price range: Customer A: Product type interest = sneakers; Brand interest = Nike; Price range interest = 300-500 yuan; Customer B: Product type interest = sneakers; Brand interest = Nike; Price range interest = 300-500 yuan; Customer C: Product type interest = mobile phones; Brand interest = Apple; Price range interest = 5000-7000 yuan. Based on this data, Customer A and Customer B are highly similar across all interest dimensions. They both prefer sneakers (product type), Nike (brand), and the 300-500 yuan price range; therefore, they will be clustered into the same group. However, Customer C's interests are completely different, primarily preferring mobile phones (product type), Apple (brand), and the 5000-7000 yuan price range; therefore, they will be divided into a different group. After completing the initial clustering step, we can further divide the interest group into different subgroups based on the similarity of interest dimensions. Subgroups can help us gain a deeper understanding of the customer's interest levels. For example, a certain group may have a strong interest in a certain type of product, but have preferences in terms of brand or price. For example, continuing to use the customer data above, if we further analyze customer A and customer B, although they are highly similar in product type (sports shoes), brand (Nike), and price range (300-500 yuan), they may still differ in some details. For example, customer A: frequently buys Nike running shoes; customer B: likes to buy Nike basketball shoes. Based on this difference, we can further divide customer A and customer B into two subgroups: the running shoe preference subgroup and the basketball shoe preference subgroup. Although they belong to the same interest group (sports shoe preference group) in terms of overall interest, they differ in product subcategories, thus forming different subgroups. To better understand this process, cluster analysis can be broken down into the following steps: Data preparation: Collect interest dimension data for each customer, including product type, brand, price range, etc.; Calculate similarity: Use similarity measures (such as Euclidean distance or cosine similarity) to calculate the similarity of customers across different interest dimensions; Clustering algorithm: Use clustering algorithms (such as K-means or hierarchical clustering) to group similar customers together; Analyze group characteristics: Determine the core interests and preferences of each group by analyzing its interest dimensions; Subgroup segmentation: Based on the detailed differences in interest dimensions, further subdivide the large group into multiple interest subgroups. After completing the cluster analysis, personalized marketing strategies and recommendation systems can be developed based on each customer's interest group and subgroup. For example, for the group that prefers athletic shoes, Nike athletic shoes can be recommended, while for the subgroup that prefers running shoes, the latest running shoe models can be prioritized. For customers who like Apple phones, Apple phones in different price ranges can be recommended based on their price range preferences.

[0029] In an optional embodiment, the method further includes: In response to the detection of customer feedback data on the first set of marketing resources pushed to the system, a second set of customer interest preferences is obtained based on the feedback data; wherein, the feedback data includes the customer's click-through rate, dwell time and enthusiasm for participating in marketing activities.

[0030] It's important to note that when the first batch of marketing resources (such as advertisements, recommended products, and promotional information) is pushed to customers, their behavior (e.g., clicks, views, and engagement) provides valuable feedback data. Based on this data, we can further determine customers' secondary interest preferences, thereby better understanding their actual needs and interests. Specific feedback data includes: Click-through rate (CTR): the number and frequency of times customers click on marketing content; a high CTR usually indicates customer interest in the content, while a low CTR suggests the content may not meet customer needs; Dwell time: the time customers spend on the marketing resource page; a longer dwell time indicates customer interest in the content and a potential for further exploration of related products or information; Engagement with marketing activities: whether customers participate in marketing activities, such as filling out forms, participating in promotions, registering for memberships, etc., and the frequency and extent of participation; active participation indicates a high level of interest and acceptance of the marketing activities and content. By analyzing customer behavior after receiving marketing resources, we can obtain their secondary interest preferences; for example, customers may show continued interest in similar products after clicking on a particular type of product advertisement, or they may demonstrate a high level of interest in a particular brand when participating in its promotional activities. For example: Suppose we push some sneaker ads to a group of customers, including sneakers from different brands (Nike, Adidas, Puma, etc.) and at different price ranges (low, mid, and high). Customer feedback data is as follows: Customer A: Clicked on a Nike sneaker ad; stayed on the ad page for 5 minutes, viewed details; participated in Nike's promotional activities, and registered as a member; Customer B: Clicked on an Adidas sneaker ad; stayed on the ad page for only 1 minute, did not view details, and did not participate in the promotional activities. Based on this data, we can draw the following conclusions: Customer A is very interested in Nike sneakers, not only clicking the ad but also spending time viewing details and participating in the promotional activities. Therefore, their secondary interest preference characteristic can be updated to confirm their preference for Nike sneakers; Customer B, although clicking on the Adidas ad, had a short dwell time and did not participate in the activities, indicating a lower interest in that brand. Their secondary interest preference characteristic may lean towards other brands or other product types. Feedback data not only identifies changes in customer interests but also allows for adjustments to marketing strategies to better align with customers' latest needs. For example, if a customer shows a strong interest in Nike sneakers, more related Nike products or sneakers can be recommended, or even customized offers can be provided. For instance, suppose a customer (Customer C) shows great interest in high-end sneakers (such as limited-edition Adidas models), clicking and staying on the feed for a considerable time after the initial marketing push. In this case, based on their interest in high-end sneakers, more limited-edition Adidas models or similar high-end sports brands can be recommended. Alternatively, more moderately priced mid-to-high-end shoes can be offered, or exclusive discounts can be provided to further stimulate customer interest. This continuous feedback analysis helps to build more accurate customer profiles and understand changes in customer interests over different time periods. Customers' secondary interest preferences may constantly adjust and update with their behavior; therefore, marketing strategies also need to be dynamically optimized in response to changes in customer interests. For example, Customer D: Upon receiving the first marketing push, the customer clicked on an ad for running shoes in the low-price range, but only stayed for a short time. However, later during the promotion, the customer purchased a pair of mid-priced running shoes. Analysis shows that the customer's secondary interest preferences indicate a potential interest in high-performance running shoes. Subsequent pushes should focus more on promoting such products and provide more information about the shoes' effectiveness and value.

[0031] In an optional embodiment, a matching degree assessment is performed on the first marketing resource set and the interest group category to obtain a matching degree assessment result, including: Based on the historical feedback data of customers in the interest group category to marketing resources, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of the interest group category, the results are compared with the preset matching threshold to determine the matching evaluation result. Specifically, if the proportion of positive feedback in the historical feedback data of customers in the interest group category to marketing resources is greater than the preset first matching degree threshold, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of the interest group category is greater than the preset second matching degree threshold, then the matching degree assessment result is judged to be a high matching degree.

[0032] It's important to clarify the following: Customer historical feedback data: This is based on customer feedback to certain marketing resources (such as advertisements, recommendations, promotions, etc.), especially positive feedback such as clicks, purchases, and participation. The relevance of marketing resources to the target audience: This refers to whether the pushed marketing resources (such as a product or advertisement) match the preferences and interests of the target audience. The matching threshold: This is a pre-set standard used to determine the degree of matching between resources and the target audience. The threshold is typically used to ensure that the relevance between resources and customer needs reaches a certain level. Matching is determined through two conditions: the proportion of positive feedback refers to customers' positive responses to a marketing resource, such as clicks, registrations, and purchases; if the proportion of positive feedback from customers within the target audience to a marketing resource is greater than a pre-set first matching threshold, it indicates that the group has a high interest in that type of resource. Relevance refers to the degree to which the pushed marketing resources match the interests (such as preferences and needs) of the customer's interest group. This relevance is usually assessed based on the similarity of the customer group's historical behavior, interest categories, and marketing content. If the relevance between the marketing resources in the first set of marketing resources and the interests of the interest group is greater than the second relevance threshold, it means that the resource is suitable for this group. When both conditions are met, it can be judged as a high relevance, that is, the marketing resource is relevant and effective for the interest group and can be pushed to them. Conversely, if the conditions are not met, it may not be suitable to push the resource. Specific example: Suppose a marketing campaign is underway targeting sports enthusiasts, with a set of marketing resources, such as advertisements for athletic shoes and coupons for gym equipment, and the goal is to push these resources to the target audience. The following is a specific example to illustrate the evaluation process: Marketing Resources: Sneaker ads; Gym equipment coupons; Healthy eating recommendations; Interest Group Category: Sports enthusiasts: including people who enjoy running, fitness, basketball, etc.; First Matching Threshold: Set at 60%, meaning that only when more than 60% of customers in the interest group give positive feedback on a certain type of marketing resource can it be considered to meet this condition; Second Matching Threshold: Set at 75%, meaning that only when the marketing resource matches the interests of the interest group by more than 75% can it be considered to meet this condition. Customer A: High click-through rate on the sneaker ad, made several purchases, positive feedback rate 80%; Customer B: High click-through rate on the fitness equipment coupon, but did not purchase, positive feedback rate 55%; Customer C: Indifferent response to the healthy eating recommendation ad, positive feedback rate 40%; Based on the first matching threshold (60%), it can be considered that Customer A's feedback on the sneaker ad meets the standard, but the feedback ratios of Customers B and C do not reach the preset standard, therefore they do not meet the condition; The relevance of athletic shoe ads to sports enthusiasts is 80% (high relevance, as sports enthusiasts typically have a strong demand for athletic shoes); the relevance of gym equipment coupons to sports enthusiasts is 70% (relatively high relevance, but slightly lower than athletic shoe ads); the relevance of healthy eating recommendations to sports enthusiasts is 50% (low relevance, as healthy eating, while relevant, is not a core interest of the group); based on the second matching threshold (75%), neither athletic shoe ads nor gym equipment coupons meet the 75% relevance standard, only athletic shoe ads meet the criteria. Sneaker ads: With a positive feedback rate of 80% (greater than 60%) and a relevance of 80% (greater than 75%), sneaker ads are judged as having a high relevance. Gym equipment coupons: With a positive feedback rate of 55% (less than 60%), although the relevance is relatively high, the feedback rate does not meet the first condition, therefore it does not meet the high relevance requirement. Healthy diet recommendations: Both the positive feedback rate and the relevance do not meet the threshold requirements, therefore they are rated as having a low relevance.

[0033] In an optional embodiment, the method further includes: In response to the lack of detected customer feedback data on the first set of marketing resources pushed to the system, supplementary customer interaction data is obtained, including customer behavior data on other relevant platforms. Based on supplementary interaction data, the customer's supplementary interest preference characteristics are determined, and the interest group categories are updated based on the supplementary interest preference characteristics to obtain the updated interest group categories; Based on the updated interest group categories, a second set of marketing resources is selected from the marketing resource library; A new matching assessment is conducted on the second set of marketing resources and the updated interest group categories; If the new matching assessment results in a high match, the second set of marketing resources will be pushed to customers in the corresponding updated interest group category.

[0034] It's important to note that when customers don't provide feedback on certain marketing resources—for example, no clicks, no purchases, no engagement—the system will collect supplementary interaction data. This typically refers to customer behavior data on other platforms; for example: likes, comments, and shares on social media platforms; browsing history and purchase records on other e-commerce platforms; and customer search keywords in search engines. This supplementary data helps the system gain a more comprehensive understanding of customer interests and preferences. Even if there's no direct response on a particular marketing platform, it can still be supplemented by data from other platforms. By analyzing this supplementary interaction data, the system can extract customers' supplementary interest preferences. For example, if customers frequently browse product pages for certain sports equipment or running shoes, or like sports-related content on social media, the system can infer that they have a strong interest in sports equipment or running products. If customers frequently search for "fitness plans" or "dietary weight loss" online, it can infer that they have a high interest in healthy eating and fitness plans. Based on the customer interest preferences obtained from the supplementary interaction data, the system will update the original interest group categories. For example, assuming that the original "sports enthusiasts" group's interests were mainly focused on running and fitness equipment, but through supplementary interaction data the system found that some customers also had a strong interest in "yoga," then the system will add yoga as a new interest to this group, thereby updating the interest group category. After updating the interest group categories, the system will re-select a new batch of marketing resources from the marketing resource library that better match the needs of the updated interest groups. This is the second marketing resource set. For example, if the "sports enthusiasts" group updates its interest in "yoga" products, the system may select "yoga mats" or "yoga class discounts" as new marketing resources. If some members of the group are still interested in "running shoes," the system may retain the previous "running shoe ads" as one of the marketing resources. Based on the updated interest group categories, the system will conduct a new matching evaluation on the selected second marketing resource set. The evaluation method is similar to before: checking whether the resource matches the customer's preferences and comparing it with a preset matching threshold. If the second marketing resource set has a high matching degree with the updated interest group categories and exceeds the preset matching threshold, then the resource set is considered to have a high matching degree. Finally, when the new matching evaluation results show that certain marketing resources are highly compatible with the updated interest group categories, these marketing resources will be pushed to the corresponding customers. At this time, the marketing resources received by customers are more likely to arouse their interest and engagement. A concrete example: Suppose there's a group of "sports enthusiasts" whose original interests were "running" and "fitness equipment." Supplementing their interaction data reveals the following: Customer A frequently likes content related to "yoga" on social media; Customer B purchased yoga mats and yoga classes on other e-commerce platforms; Customer C searched for and browsed yoga-related products on multiple platforms. Based on this data, the system determines that "yoga" has become a new interest for this group and updates its interest preferences to include "yoga." Then, the system selects a second set of marketing resources based on the updated interest group category: for example, pushing "yoga mats" and "yoga class discounts." Simultaneously, running shoe ads and gym equipment coupons continued to be pushed, based on customers' long-term interest in this content. Finally, a new matching assessment was conducted, with the following results: For "yoga mats" and "yoga class discounts," the matching assessment showed that these marketing resources highly aligned with the new needs of the interest group, and were therefore considered to have a high matching degree. For running shoe ads and gym equipment coupons, although there was still some matching degree, the matching degree of these resources may be slightly lower due to the lack of new "yoga" interest points. Ultimately, the marketing resources for yoga mats and yoga class discounts were pushed to the updated interest group to ensure that these customers received content that better matched their current interests, thereby improving marketing effectiveness.

[0035] In an optional embodiment, the matching degree evaluation of the first marketing resource set and interest group categories to obtain the matching degree evaluation result further includes: If, within the interest group category, the proportion of positive feedback from customers in each interest subgroup regarding marketing resources in historical feedback data is greater than the preset first matching threshold, and the degree of fit between each marketing resource in the first marketing resource set and the interest points of each interest subgroup is greater than the preset second matching threshold, then the matching degree assessment result is judged to be a high matching degree.

[0036] It should be noted that, firstly, interest groups can be further divided into interest subgroups; for example, under the broad category of "sports enthusiasts," there may be multiple interest subgroups, such as: running enthusiasts; fitness equipment users; yoga enthusiasts. These subgroups may have different interests and responses to marketing resources, so they need to be evaluated separately. The first marketing resource set refers to the collection of all marketing resources selected from the marketing resource pool; for example: running shoe advertisements; fitness equipment promotions; yoga class discounts; these marketing resources are designed for the "sports enthusiast" group, but may not be entirely suitable for all subgroups; Historically, feedback from each interest subgroup to a marketing resource can be categorized into positive feedback (e.g., clicks, purchases, shares) and negative feedback (e.g., ignoring, unsubscribing). When evaluating relevance, if the proportion of positive feedback within each interest subgroup exceeds a preset first relevance threshold, it indicates a high level of interest in the resource from that subgroup, warranting further consideration. For example, assuming the "running enthusiasts" subgroup has a 70% positive feedback rate on running shoe ads, and the preset first relevance threshold is 50%, it meets the criteria. Similarly, if the "yoga enthusiasts" subgroup has an 80% positive feedback rate on yoga class discounts, and the threshold remains at 50%, it also meets the criteria. Next, the relevance of each resource and interest subgroup in the first marketing resource set is evaluated. The relevance of each marketing resource and interest point is calculated by analyzing the degree of matching between marketing content and customer preferences. If the relevance of each marketing resource and each interest subgroup is greater than the preset second matching threshold, it indicates that the content of these resources is highly relevant to the subgroup and can arouse their interest. Specific relevance evaluations include: relevance evaluation of running shoe advertisements and running enthusiasts subgroup: assuming a relevance of 85%, if the preset second matching threshold is 70%, it meets the condition; relevance evaluation of yoga course discounts and yoga enthusiasts subgroup: assuming a relevance of 90%, if the threshold is 80%, it also meets the condition. If the proportion of positive feedback from each interest subgroup is greater than the first matching threshold, and the relevance of each resource in the first marketing resource set to each interest subgroup is greater than the second matching threshold, then the relevance between this set of resources and groups can be judged as a high match. This means that these marketing resources are highly relevant to this group or subgroup, and customers are more likely to respond positively to them, so they can be pushed to this group. Specific example: Suppose there is a "sports enthusiast" group, which includes three subgroups: running enthusiasts, fitness equipment users, and yoga enthusiasts; marketing resources include running shoe advertisements, fitness equipment promotions, and yoga class discounts; next, the relevance of these resources is evaluated: Running enthusiast subgroup: Positive feedback rate for running shoe advertisements: 70% (higher than the preset first relevance threshold of 50%); Relevance of running shoe advertisements: 85% (higher than the preset second relevance threshold of 70%); Fitness equipment user subgroup: Positive feedback rate for fitness equipment promotions: 60% (higher than the first relevance threshold of 70%). The matching degree thresholds are as follows: 1) 50% for running shoe advertising, 75% for fitness equipment promotions (higher than the second matching degree threshold of 70%), and 80% for yoga enthusiasts (higher than the first matching degree threshold of 50%). The matching degree for yoga class discounts is 90% (higher than the second matching degree threshold of 80%). Based on these evaluation results, the matching degree of running shoe advertising, fitness equipment promotions, and yoga class discounts all meet the two conditions of "positive feedback ratio" and "matching degree". Therefore, it can be considered that these marketing resources have a high matching degree with each subgroup and can be pushed to the corresponding customers.

[0037] In an optional embodiment, the matching degree evaluation of the first marketing resource set and interest group categories to obtain the matching degree evaluation result further includes: If, within an interest group category, the proportion of positive feedback from customers within a subgroup regarding marketing resources is less than the first matching threshold, the matching assessment result is determined to be a low match, and marketing resources for the subgroup are re-screened; or... If the degree of fit between a marketing resource in the first set of marketing resources and the interest points of an interest subgroup is less than the second matching threshold, then the matching degree assessment result is judged as low matching degree, and the marketing resources are adjusted or replaced.

[0038] It should be noted that under an interest group category (such as "sports enthusiasts"), there may be multiple interest subgroups, such as running enthusiasts, fitness equipment users, and yoga enthusiasts. Different subgroups may respond differently to marketing resources, so each subgroup needs to be evaluated separately. The first set of marketing resources refers to the marketing resources selected for a specific group or subgroup; for example: running shoe advertisements; fitness equipment promotions; yoga class discounts. These marketing resources may have taken into account the needs of different subgroups when they were designed, but not every subgroup will be interested in every resource. When conducting a matching evaluation, two main criteria must be met: First, the historical feedback data of each interest subgroup on marketing resources can be divided into positive and negative feedback. The system calculates the positive feedback ratio, i.e., the proportion of positive feedback to all feedback. This is then compared with a preset first matching threshold. If the positive feedback ratio of a subgroup to a particular marketing resource is lower than the threshold, it indicates that the resource is not very attractive to that subgroup, and adjustments or re-screening of resources should be considered. Second, there must be a certain degree of fit between the marketing resource and the interests of the interest subgroup, reflecting the relevance of the resource content to the subgroup's needs. If the fit between a marketing resource and the interests of a particular interest subgroup is lower than a preset second matching threshold, it means that the resource is not relevant enough to that group and should be adjusted or replaced. If any of the following situations are found during the evaluation process, the matching degree is considered low: Positive feedback ratio lower than the first matching threshold: This means that the feedback from a subgroup on the marketing resource is unsatisfactory, and the system will re-screen marketing resources suitable for that subgroup based on historical feedback. Fit lower than the second matching threshold: This means that a marketing resource does not match the interests of a particular subgroup, and the system will re-evaluate the effectiveness of the resource and may adjust or replace it. Specific example: Suppose there is a "sports enthusiast" group, which includes three subgroups: running enthusiasts, fitness equipment users, and yoga enthusiasts; marketing resources include: running shoe advertisements, fitness equipment promotions, and yoga class discounts; Subgroup 1: Running enthusiasts; Positive feedback rate for running shoe advertisements: 65% (above the first matching threshold of 50%); The relevance of running shoe advertisements to running enthusiasts: 80% (above the second matching threshold of 70%); In this subgroup, both the feedback and relevance of running shoe advertisements meet the requirements, therefore the matching degree is high; Subgroup 2: Fitness equipment users; Positive feedback rate for fitness equipment promotions: 40% (below the first matching threshold of 50%); Fitness equipment promotions and fitness equipment... User fit: 85% (above the second matching threshold of 70%); In this subgroup, the positive feedback rate for fitness equipment promotions is below the threshold, so the system will judge the fit as low; At this time, the system will re-select other marketing resources suitable for fitness equipment users, such as gym membership card discounts or new fitness equipment promotions, instead of continuing to use the current fitness equipment promotions; Subgroup 3: Yoga enthusiasts; Positive feedback rate for yoga class discounts: 80% (above the first matching threshold of 50%); Fit between yoga class discounts and yoga enthusiasts: 75% (above the second matching threshold of 70%); In this subgroup, both the feedback and fit of yoga class discounts meet the requirements, so the fit is high.

[0039] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A back office operations management system based on big data, characterized in that, The method comprises the following steps: a data analysis unit is used to obtain first interaction data of a customer, determine first interest preference features of the customer based on the first interaction data; wherein the first interaction data comprises browsing records, search records and purchase records of the customer on a marketing platform; a group classification unit is used to perform cluster analysis on the first interest preference features, and determine interest group categories of the customer; wherein the interest group categories are divided according to the similarity of customer interests; a resource screening unit is used to screen a first marketing resource set from a marketing resource library based on the interest group categories; wherein the first marketing resource set comprises marketing content, marketing activity schemes and marketing preferential information related to the interest group categories; a resource matching unit is used to perform matching degree evaluation on the first marketing resource set and the interest group categories, and obtain a matching degree evaluation result; an operation management unit is used to push the first marketing resource set to customers of a corresponding interest group category in response to the matching degree evaluation result representing high matching degree.

2. The big data based back office operations management system of claim 1, wherein, Determination of first interest preference features of the customer based on the first interaction data comprises: extracting a plurality of basic features from the first interaction data, performing association analysis on the plurality of basic features, and obtaining an association feature combination; wherein the basic features comprise browsing product categories, search keywords and purchase product price intervals; performing cluster analysis on the customer based on the association feature combination, and obtaining a plurality of interest groups; each interest group comprises customers with similar interest features; for each interest group, analyzing the corresponding association feature combination to determine the first interest preference features corresponding to each interest group, and obtaining the first interest preference features of each customer.

3. The big data based back office operations management system of claim 2, wherein, The association analysis on the plurality of basic features to obtain the association feature combination comprises: analyzing the semantic association between browsing product categories and search keywords to determine a browsing product category and search keyword combination with semantic consistency; analyzing the price adaptation relationship between purchase product price intervals and browsing product categories to determine a purchase product price interval and browsing product category combination with price adaptation; fusing the combination with semantic consistency and the combination with price adaptation to obtain the association feature combination; analyzing the browsing sequence of different product categories in the browsing records to determine the preference sequence of the customer in browsing products; associating the preference sequence of browsing products with browsing product categories to obtain an association feature combination comprising the browsing sequence.

4. The big data based back office operations management system of claim 3, wherein, For each interest group, analyzing the corresponding association feature combination to determine the first interest preference features corresponding to each interest group, and obtaining the first interest preference features of each customer, comprises: for each interest group, counting the occurrence frequency of each basic feature in the association feature combination, and determining the basic feature with the highest occurrence frequency as the main interest feature of the interest group; based on the main interest feature and other features in the association feature combination, determining the first interest preference features corresponding to the interest group; according to the interest group to which the customer belongs, assigning the first interest preference features corresponding to the interest group to the customer to obtain the first interest preference features of each customer; For each interest group, analyze the correlation strength between different features in the associated feature combination, and determine the hierarchy of interest preferences according to the correlation strength; Based on the hierarchy of interest preferences, refine the first interest preference feature corresponding to the interest group.

5. The big data based back office operations management system of claim 4, wherein, The first interest preference feature includes multiple interest dimensions, the interest group category includes multiple interest subgroups, and the interest dimensions include product type interest, brand interest, and price interval interest; The clustering analysis of the first interest preference feature determines the interest group category of the customer, including: In the multiple interest dimensions, the customers with similar product type interest, similar brand interest, and similar price interval interest are determined as a group; Based on the similarity of each group of customers in each interest dimension, the interest subgroup is determined.

6. The big data based back office operations management system of claim 5, wherein, The method further includes: In response to detecting feedback data of the customer to the pushed first marketing resource set, obtaining the second interest preference feature of the customer based on the feedback data; wherein the feedback data includes the click rate, the stay time and the enthusiasm of the customer to participate in the marketing activity.

7. The big data based back office operations management system of claim 6, wherein, The matching degree evaluation of the first marketing resource set and the interest group category obtains the matching degree evaluation result, including: Based on the historical feedback data of the customer to the marketing resource in the interest group category, and the fit degree of each marketing resource in the first marketing resource set and the interest point of the interest group category, respectively compared with the preset matching degree threshold, to determine the matching degree evaluation result; Wherein, when the proportion of positive feedback in the historical feedback data of the customer to the marketing resource in the interest group category is greater than the preset first matching degree threshold, and when the fit degree of each marketing resource in the first marketing resource set and the interest point of the interest group category is greater than the preset second matching degree threshold, it is judged that the result of the matching degree evaluation is high matching degree.

8. The big data based back office operations management system of claim 7, wherein, The method further includes: In response to not detecting feedback data of the customer to the pushed first marketing resource set, obtaining the supplementary interaction data of the customer, wherein the supplementary interaction data includes the behavior data of the customer on other related platforms; Based on the supplementary interaction data, determine the supplementary interest preference feature of the customer, update the interest group category based on the supplementary interest preference feature, and obtain the updated interest group category; Based on the updated interest group category, reselect the second marketing resource set from the marketing resource library; New matching degree evaluation is performed on the second marketing resource set and the updated interest group category; In response to determining that the result of the new matching degree evaluation is high matching degree, the second marketing resource set is pushed to the customers corresponding to the updated interest group category. 9.The big data-based back office management system of claim 8, wherein, The matching degree evaluation of the first marketing resource set and the interest group category obtains the matching degree evaluation result, further including: If in the interest group category, the proportion of positive feedback in the historical feedback data of the customers in each interest subgroup to the marketing resource is greater than the preset first matching degree threshold, and the matching degree of each marketing resource in the first marketing resource set to the interest point of each interest subgroup is greater than the preset second matching degree threshold, it is judged that the result of the matching degree evaluation is high matching degree. 10.The big data-based back office management system of claim 9, wherein, The matching degree evaluation on the first marketing resource set and the interest group category is performed to obtain a matching degree evaluation result, and the matching degree evaluation result further includes: If in the interest group category, the proportion of positive feedback in the historical feedback data of the customers in each interest subgroup to the marketing resource is less than the first matching degree threshold, it is judged that the result of the matching degree evaluation is low matching degree, and the marketing resource is reselected for the interest subgroup; or, If the matching degree of each marketing resource in the first marketing resource set to the interest point of each interest subgroup is less than the second matching degree threshold, it is judged that the result of the matching degree evaluation is low matching degree, and the marketing resource is adjusted or replaced.

Images