A method and system for recommending prepared dishes

By segmenting users by type and dynamically analyzing user behavior, the system assesses the uniformity and divergence of intentions among frequent buyers of pre-prepared meals, adjusts recommendation strategies, and solves the problem of inaccurate recommendations in the existing system, achieving higher user satisfaction and market responsiveness.

CN118710375BActive Publication Date: 2024-11-05GUIZHOU FOOD ENG VOCATIONAL COLLEGE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411195550.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-05
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

Existing ready-to-eat meal recommendation systems struggle to respond promptly to changes in user preferences, resulting in inaccurate and undiversified recommendations that fail to effectively differentiate between user purchase frequency and willingness to explore new products.

Method used

By analyzing user behavior based on a database, users are categorized into frequent and infrequent buyers of pre-made meals. The uniformity and divergence of user intentions are assessed, and recommendation strategies are dynamically adjusted, including updating user profiles and assessing diverse needs. Random forest algorithms and information entropy calculations are used to adjust the breadth of recommended categories.

Benefits of technology

It improved the accuracy of pre-made meal recommendations and user satisfaction, enhanced the platform's responsiveness to market trends, optimized inventory management, reduced resource waste, and increased user loyalty and purchase motivation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118710375B_ABST
    Figure CN118710375B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for recommending prepared dishes, which specifically relate to the technical field of product recommendation, and are used to solve the existing problems of difficulty in timely responding to rapid changes in user preferences and failure to accurately recommend prepared dishes to users. The method divides users into two categories of those who frequently purchase prepared dishes and those who do not frequently purchase prepared dishes through database information, and for users who frequently purchase prepared dishes, screens out a set of intended prepared dishes and evaluates the degree of change in the characteristics of these prepared dishes; analyzes the re-update of user search results to evaluate the degree of divergence of user intentions for prepared dishes; determines the urgency of updating user portraits based on the evaluation results, and decides whether to update immediately; evaluates the intensity of diversity requirements that an e-commerce platform should consider when recommending prepared dishes by analyzing the entropy of search keywords of users who frequently purchase prepared dishes; and dynamically adjusts the recommendation strategy based on the user's exploration behavior to expand or narrow the recommended category range.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of product recommendation, and more specifically, to a method and system for recommending prepared dishes. Background Art

[0002] The importance of accurately recommending pre-prepared meals cannot be ignored. It directly affects users' purchasing decisions and satisfaction, which in turn has a significant impact on user retention and the platform's overall sales performance. Personalized recommendation systems are crucial to improving user satisfaction with pre-prepared meals and increasing sales of pre-prepared meals.

[0003] In the prepared food market, consumers' preferences and demands are extremely diverse and dynamically changing. When e-commerce platforms recommend prepared food to users, especially for users who frequently pay attention to and purchase prepared food, traditional recommendation systems often rely on static user portraits and past purchase history, making it difficult to respond to rapid changes in user preferences in a timely manner. They often cannot effectively distinguish between users' purchase frequency and willingness to explore new products, resulting in inaccurate or individuated recommendations, and inability to accurately recommend prepared food to users.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for recommending prepared dishes to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for recommending pre-prepared dishes comprises the following steps:

[0008] Obtain potential recommended users of pre-prepared dishes based on the database, and divide the potential recommended users into users who frequently purchase pre-prepared dishes and users who do not frequently purchase pre-prepared dishes;

[0009] Analyze the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes, screen out the intended pre-prepared dish set, analyze the degree of change in the characteristics of the intended pre-prepared dish set, and evaluate the uniformity of the intended pre-prepared dish of users who frequently purchase pre-prepared dishes;

[0010] Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals;

[0011] Based on the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals and the divergence of their pre-prepared meal intention, the urgency of updating the user profile of the users who frequently purchase pre-prepared meals is evaluated, and it is determined whether to update the user profile immediately.

[0012] By analyzing the entropy of keywords searched by users who frequently purchase prepared meals, the intensity of demand for diversity in the recommendations of prepared meals on e-commerce platforms is evaluated;

[0013] By conducting a comprehensive analysis of the uniformity of pre-prepared meal intentions among users who frequently purchase pre-prepared meals, the degree of divergence of pre-prepared meal intentions among users who frequently purchase pre-prepared meals, and the intensity of demand for diversity in pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms are adjusted.

[0014] In a preferred embodiment, potential recommended users of pre-prepared dishes are obtained based on the database, and the potential recommended users are divided into users who often purchase pre-prepared dishes and users who do not often purchase pre-prepared dishes, specifically:

[0015] The data collected in the database includes basic user information, transaction information, browsing history, and product information;

[0016] Construct features for user classification based on the data collected from the database. The features used for user classification include purchase frequency, purchase preference, consumption level, and activity level;

[0017] Label users according to purchase frequency, divide the data into training and test sets, and use the random forest algorithm to train the model;

[0018] The trained model is used for user classification, and potential recommendation users are divided into users who frequently purchase pre-prepared meals and users who do not frequently purchase pre-prepared meals.

[0019] In a preferred embodiment, the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes are analyzed, and a set of intended pre-prepared dishes is obtained by screening. The degree of change in the characteristics of the set of intended pre-prepared dishes is analyzed, and the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes is evaluated, specifically:

[0020] S101: establishing a desired pre-prepared meal set by identifying frequently purchased and browsed pre-prepared meal products based on historical purchase data and browsing behavior of users who frequently purchase pre-prepared meals;

[0021] S102: Collecting transaction records, user reviews, browsing data and other interactive data of the intended pre-prepared meal set;

[0022] S103: extracting key features for each intended pre-prepared dish in the intended pre-prepared dish set, where the key features include convenience, category, and homogeneity;

[0023] S104: Calculate the standard deviation of each key feature, and quantify the volatility of the key feature change according to the standard deviation calculation result of the key feature;

[0024] S105: assigning a weight to each key feature, performing weighted processing on the standard deviation of the key features and then summing them up, obtaining a comprehensive change index of intention for the intended pre-prepared meal set, and evaluating the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals based on the comprehensive change index of intention;

[0025] The calculation formula of the intention comprehensive change index is: ;

[0026] in, is the comprehensive change index of intention, is the number of key features, is the key feature number, It is The weight of the key features, It is The standard deviation calculation results corresponding to the key features.

[0027] In a preferred embodiment, the re-update of the search results of users who frequently purchase pre-prepared dishes is analyzed to evaluate the divergence of the pre-prepared dishes intentions of users who frequently purchase pre-prepared dishes, specifically:

[0028] The process from when users who frequently purchase pre-prepared food start searching for pre-prepared food to when they exit the e-commerce platform is marked as a search task;

[0029] Establish a search task set, which includes multiple search tasks of users who frequently purchase pre-prepared dishes recently;

[0030] Determine whether there is a re-search situation in the search task, and mark the search task with the re-search situation in the search task set as a divergent search task;

[0031] The ratio of the number of divergent search tasks in the search task set to the total number of search tasks in the search task set is marked as the divergent search ratio;

[0032] Record the total number of searches conducted by users who frequently purchase pre-prepared meals in the divergent search task;

[0033] For each search in the divergent search task, record the browsing time of users who frequently purchase pre-prepared meals on the search results page;

[0034] Calculate the re-search browsing time ratio;

[0035] The search intention divergence value of each divergent search task is calculated by assigning weights to the re-search browsing time ratios in the divergent search task and summing them up;

[0036] Sum the search intention divergence values ​​of all divergent search tasks in the search task set to obtain the set search intention divergence value;

[0037] The divergent search ratio and the set search intention divergence value are dimensionless, and the divergent search ratio after dimensionless processing and the set search intention divergence value are weighted summed to calculate the search intention divergence index, which is expressed as follows:

[0038] ;

[0039] in, is the search intention divergence index, are the divergent search ratio and the set search intention divergence value after dimensionless processing, are the weights of the divergent search ratio and the divergent value of the set search intention after dimensionless processing, and Both are greater than 0.

[0040] In a preferred embodiment, based on the uniformity of the pre-prepared food intention of the pre-prepared food frequent purchasing users and the divergence of the pre-prepared food intention of the pre-prepared food frequent purchasing users, the urgency of updating the user profile of the pre-prepared food frequent purchasing users is evaluated, and it is determined whether to update the user profile immediately, specifically:

[0041] When the comprehensive change index of intention is less than or equal to the comprehensive change threshold of intention, and the search intention divergence index is less than or equal to the search intention divergence threshold, it is determined not to update the user portrait immediately; otherwise, it is determined to update the user portrait immediately;

[0042] When it is determined that the user portrait is not to be updated immediately, the user portrait is updated based on a preset update frequency.

[0043] In a preferred embodiment, the intensity of demand for diversity in the pre-prepared meal recommendations on the e-commerce platform is evaluated by analyzing the entropy of keywords searched by users who frequently purchase pre-prepared meals, specifically:

[0044] S201: Collect all search queries for pre-prepared dishes conducted by users who frequently purchase pre-prepared dishes on the e-commerce platform within a preset period; the search queries include keywords entered by users who frequently purchase pre-prepared dishes, search frequencies, and search dates;

[0045] S202: extract all independent keywords; sort and clean the keywords and remove meaningless filler words;

[0046] S203: Count the number of times each keyword appears in all search queries, and then calculate the frequency of occurrence of each keyword; the frequency of occurrence calculation formula of each keyword is: ;in, For the The frequency of occurrence of keywords, For the The number of times a keyword appears, is the total number of occurrences of all keywords;

[0047] S204: Apply the information entropy formula to calculate the search diversity entropy evaluation index, which is expressed as: ; is the search diversity entropy evaluation index, The total number of unique keywords.

[0048] In a preferred embodiment, by comprehensively analyzing the homogeneity of the pre-prepared food intention of users who frequently purchase pre-prepared food, the divergence of the pre-prepared food intention of users who frequently purchase pre-prepared food, and the intensity of the demand for diversity of pre-prepared food recommendations on the e-commerce platform, the breadth of categories of pre-prepared food recommended by the e-commerce platform to the users who frequently purchase pre-prepared food is adjusted, specifically:

[0049] The intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are normalized, and the normalized intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are respectively assigned preset proportional coefficients to calculate the recommendation extensive evaluation coefficient, which is expressed as follows:

[0050] ;

[0051] in, To recommend a broad assessment coefficient, is the base of natural logarithm, are the preset proportional coefficients of the intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index, respectively, and All are greater than 0;

[0052] The relationship between the recommendation breadth evaluation coefficient and the breadth of categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals is: the larger the recommendation breadth evaluation coefficient, the greater the breadth of categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals should be adjusted.

[0053] On the other hand, the present invention provides a prepared food recommendation system, including a user segmentation module, an intention uniformity evaluation module, an intention divergence evaluation module, a portrait update judgment module, a recommendation diversity evaluation module, and a recommendation extensive adjustment module;

[0054] User segmentation module: obtains potential recommended users of pre-prepared dishes based on the database, and divides the potential recommended users into users who frequently purchase pre-prepared dishes and users who rarely purchase pre-prepared dishes;

[0055] Intention uniformity evaluation module: Analyze the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes, screen out the intended pre-prepared dishes set, analyze the degree of change in the characteristics of the intended pre-prepared dishes set, and evaluate the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes;

[0056] Intention divergence evaluation module: Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of intention divergence of users who frequently purchase pre-prepared meals;

[0057] Profile update judgment module: Based on the uniformity of the pre-prepared food intention of the pre-prepared food frequent purchasing users and the divergence of the pre-prepared food intention of the pre-prepared food frequent purchasing users, the urgency of updating the user profile of the pre-prepared food frequent purchasing users is evaluated, and it is judged whether to update the user profile immediately;

[0058] Recommendation diversity evaluation module: By analyzing the entropy of search keywords of users who frequently purchase pre-prepared dishes, the diversity demand intensity of pre-prepared dishes recommended by e-commerce platforms is evaluated;

[0059] Recommendation breadth adjustment module: Through a comprehensive analysis of the uniformity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, and the intensity of demand for diversity of pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms is adjusted.

[0060] The technical effects and advantages of the method and system for recommending prepared dishes of the present invention are as follows:

[0061] 1. Based on the database information, users are divided into two categories: those who frequently purchase pre-prepared dishes and those who do not frequently purchase them. For users who frequently purchase them, their intended pre-prepared dishes are further analyzed, including screening out the intended pre-prepared dishes set and evaluating the degree of change in the characteristics of these pre-prepared dishes. In addition, the re-update of user search results is analyzed to evaluate the degree of divergence of users' intentions for pre-prepared dishes. This step understands the uniformity and diversity needs of users in the selection of pre-prepared dishes; finally, the urgency of updating the user portrait is judged based on the evaluation results, and it is decided whether to update it immediately to ensure that the recommended content is consistent with the user's current preferences. By analyzing the entropy of the search keywords of users who frequently purchase pre-prepared dishes, the intensity of diversity needs that e-commerce platforms should consider when recommending pre-prepared dishes is evaluated. The entropy of search keywords reflects the breadth of users' exploration when searching for pre-prepared dishes. It can dynamically adjust the recommendation strategy according to the user's exploration behavior and expand or narrow the recommended category range. This method not only improves user satisfaction, but also enhances the platform's responsiveness to new market trends.

[0062] 2. Significantly improved the accuracy of pre-prepared meal recommendations and user satisfaction. By dynamically analyzing users' purchase and search behaviors and updating user portraits in a timely manner, e-commerce platforms are able to provide recommendations that are more in line with users' current preferences, which directly enhances users' purchasing motivation and trust. At the same time, the implementation of this method helps e-commerce platforms to more effectively predict market trends, optimize inventory management, and reduce resource waste. In addition, by adjusting the breadth of recommended categories, the platform can better meet the needs of different user groups, especially those seeking new tastes and new experiences, thereby attracting and retaining more users. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic diagram of a recommended method for preparing dishes according to the present invention;

[0064] Figure 2 The present invention is a schematic diagram of the structure of a pre-prepared meal recommendation system. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Example 1

[0067] Figure 1 The present invention provides a method for recommending prepared dishes, which comprises the following steps:

[0068] Based on the database, potential recommended users of pre-prepared dishes are obtained, and the potential recommended users are divided into users who frequently purchase pre-prepared dishes and users who do not frequently purchase pre-prepared dishes.

[0069] The intended pre-prepared dishes of users who frequently purchase pre-prepared dishes are analyzed, and a set of intended pre-prepared dishes is obtained by screening. The degree of change in the characteristics of the set of intended pre-prepared dishes is analyzed, and the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes is evaluated.

[0070] Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of divergence of their intention for pre-prepared meals.

[0071] Based on the uniformity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals and the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the urgency of updating the user profiles of users who frequently purchase pre-prepared meals is evaluated, and a decision is made as to whether the user profiles should be updated immediately.

[0072] By analyzing the entropy of search keywords of users who frequently purchase pre-prepared meals, the intensity of demand for diversity in pre-prepared meal recommendations on e-commerce platforms is evaluated.

[0073] By conducting a comprehensive analysis of the uniformity of pre-prepared meal intentions among users who frequently purchase pre-prepared meals, the degree of divergence of pre-prepared meal intentions among users who frequently purchase pre-prepared meals, and the intensity of demand for diversity in pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms are adjusted.

[0074] Based on the database, potential recommended users of pre-prepared dishes are obtained, and the potential recommended users are divided into users who often purchase pre-prepared dishes and users who do not often purchase pre-prepared dishes. Specifically:

[0075] The database is a comprehensive database that contains user purchase history, browsing history, user basic information, transaction information and product information. The data collection of the database includes the following aspects:

[0076] User basic information: including but not limited to user ID, age, gender, geographic location, registration date, etc.

[0077] Transaction information: all transaction records, including transaction time, purchased goods, quantity, price, payment method and transaction status.

[0078] Browsing history: user’s browsing behavior data on the platform, such as browsed products, browsing time, page dwell time, etc.

[0079] Product information: Contains detailed description of the product, such as category ID, name, description, price, inventory, supplier information, etc.

[0080] To ensure the quality of data and the accuracy of analysis, detailed data preprocessing operations are required, including:

[0081] Data cleaning: Eliminate incomplete, erroneous or irrelevant data records, such as invalid user IDs, abnormal transaction records, etc.

[0082] Data integration: Integrate data from different sources to ensure that every user behavior can be accurately associated with the corresponding user ID and product ID.

[0083] Data formatting: uniformly format date, time and other fields to facilitate subsequent analysis.

[0084] Features for user classification are constructed based on the data collected from the database. The features for user classification include:

[0085] Purchase frequency: Calculate the frequency of each user's purchase of pre-prepared meals, such as the number of times each user purchases pre-prepared meals per month.

[0086] Purchase preference: Analyze the types of pre-prepared meals purchased by users and identify the main preferred pre-prepared meal categories.

[0087] Consumption level: Assess the user’s consumption level based on their average spending on pre-prepared meals and purchase frequency.

[0088] Activity: User activity is measured through indicators such as user login frequency and number of pages viewed.

[0089] Users can be labeled based on purchase frequency. For example, a threshold can be set, such that if a user purchases pre-made meals more than 10 times during the observation period, the user is labeled as "frequent purchase" (1), otherwise, the user is labeled as "infrequent purchase" (0).

[0090] Split the data into a training set and a test set, with a common ratio of 80% of the data used for training the model and 20% of the data used for testing the model. This can be achieved using the train_test_split function in Python's Scikit-learn library.

[0091] The model is trained using the random forest algorithm. Random forest is an ensemble algorithm that contains multiple decision trees. Its steps include:

[0092] Select random forest parameters: such as the number of trees (n_estimators), maximum depth (max_depth), etc.

[0093] Training model: Use RandomForestClassifier in Scikit-learn to train data. The model will learn how to predict whether the user is a "frequent purchaser" or "infrequent purchaser" based on the input features. This can be achieved through the following code:

[0094] from sklearn.ensemble import RandomForestClassifier

[0095] from sklearn.model_selection import train_test_split

[0096] # Assume X is feature data and y is label data

[0097] X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

[0098] rf_model = RandomForestClassifier(n_estimators=100, max_depth=10,random_state=42)

[0099] rf_model.fit(X_train, y_train)

[0100] Use the test set to evaluate the performance of the model, focusing on indicators such as accuracy, recall, and F1 score. You can use Scikit-learn's classification_report to obtain these indicators. This can be achieved through the following code:

[0101] from sklearn.metrics import classification_report

[0102] y_pred = rf_model.predict(X_test)

[0103] print(classification_report(y_test, y_pred))

[0104] The trained model is used for real-time or periodic user classification to divide potential recommendation users into users who frequently purchase pre-prepared dishes and users who infrequently purchase pre-prepared dishes.

[0105] Among them, the recommendation of pre-prepared dishes to users who frequently purchase pre-prepared dishes is usually based on user portraits; the recommendation of pre-prepared dishes to users who frequently purchase pre-prepared dishes based on user portraits is specifically as follows:

[0106] Use clustering algorithms (such as K-means or hierarchical clustering) to group users to find groups of users with similar purchasing and browsing behaviors.

[0107] Analyze the characteristics of each group and define user portraits.

[0108] Make personalized recommendations based on user portraits and select appropriate recommendation algorithms, including:

[0109] Collaborative filtering: Utilize the user-item interaction history to find similar users or items and make recommendations; it can be user-based collaborative filtering or item-based collaborative filtering.

[0110] Content recommendation: Analyze the attributes in the user profile, such as the favorite type of pre-made dishes, and recommend new products with similar attributes. Adjust the recommendation strategy based on user feedback data.

[0111] Hybrid recommendation system: combines the advantages of collaborative filtering and content recommendation to improve the accuracy and coverage of the recommendation system.

[0112] The intended pre-prepared dishes of users who frequently purchase pre-prepared dishes are analyzed, and a set of intended pre-prepared dishes is obtained by screening. The degree of change in the characteristics of the set of intended pre-prepared dishes is analyzed, and the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes is evaluated. Specifically:

[0113] S101: Based on the historical purchase data and browsing behavior of users who frequently purchase pre-prepared dishes, a set of intended pre-prepared dishes is established by identifying pre-prepared dishes that are frequently purchased and browsed:

[0114] Extract users’ historical purchase and browsing data from the e-commerce platform’s database, and use data mining techniques (such as association rule learning) to identify pre-prepared dishes that users frequently purchase or browse.

[0115] A set of intended prepared dishes is established based on the identified frequent items.

[0116] Identifying the types of pre-prepared dishes that users prefer helps to understand user needs more accurately and improve the accuracy of the recommendation system and user satisfaction.

[0117] S102: Collecting transaction records, user reviews, browsing data, and other interactive data of the intended pre-prepared meal set:

[0118] In order to fully understand the market performance and user interactions of intended pre-prepared meals, it is necessary to collect relevant transaction records, user reviews and browsing data.

[0119] Access the back-end database of the e-commerce platform to extract detailed transaction and interaction records of relevant pre-prepared dishes; integrate purchase data, user reviews, browsing data and other information by pre-prepared dish category.

[0120] S103: Extract key features for each intended pre-prepared dish in the intended pre-prepared dish set, where the key features include convenience, category, and homogeneity:

[0121] Convenience is usually related to the preparation time of the prepared dish, the preparation steps required, and the complexity of the tools required. To quantify convenience, several methods can be used:

[0122] The inverse of the preparation time of the pre-prepared dish is used to express convenience. The lower the preparation time of the pre-prepared dish, the higher the convenience.

[0123] The inverse of the number of steps required to prepare a pre-prepared dish is used to express convenience, and the fewer steps required to prepare a pre-prepared dish, the higher the convenience.

[0124] Categories can be coded according to the type of prepared dish for use in analysis.

[0125] Use One-Hot Encoding to assign a separate binary variable to each category. For example, if there are three types of pre-made dishes - pickled mustard greens and pork, sushi, and hamburgers, then pickled mustard greens and pork can be represented as [1,0,0], sushi as [0,1,0], and hamburgers as [0,0,1].

[0126] The degree of homogeneity indicates how similar a product is to other similar products on the market. This can be quantified by comparing the similarity of product characteristics.

[0127] Match prepared meals with similar products on the market by comparing key characteristics (e.g. ingredients, nutritional value, flavor).

[0128] The ratio of the number of matching features to the total number of features is marked as the degree of homogeneity. The greater the degree of homogeneity, the more similar it is to other pre-prepared dishes on the market.

[0129] S104: Calculate the standard deviation of each key feature, and quantify the volatility of the key feature changes based on the standard deviation calculation results of the key features:

[0130] A standard deviation analysis is performed on the convenience of all pre-prepared dishes in the intended pre-prepared dish set, and the standard deviation calculation result corresponding to the convenience is calculated.

[0131] Perform standard deviation analysis on the categories of all pre-prepared dishes in the intended pre-prepared dish set, and calculate the standard deviation calculation results corresponding to the categories.

[0132] A standard deviation analysis is performed on the homogeneity of all pre-prepared dishes in the intended pre-prepared dish set, and the standard deviation calculation result corresponding to the homogeneity degree is calculated.

[0133] The calculation of the standard deviation corresponding to convenience, category and degree of homogeneity is based on the existing common standard deviation, so it will not be repeated here.

[0134] The larger the calculated standard deviation corresponding to convenience, the greater the volatility of changes in convenience of pre-prepared dishes in the intended pre-prepared dish set. The larger the calculated standard deviation corresponding to category, the greater the volatility of changes in category of pre-prepared dishes in the intended pre-prepared dish set. The larger the calculated standard deviation corresponding to the degree of homogeneity, the greater the volatility of changes in the degree of homogeneity of pre-prepared dishes in the intended pre-prepared dish set.

[0135] S105: Assign a weight to each key feature, perform weighted processing on the standard deviation of the key features and then sum them up to obtain the comprehensive change index of the intended pre-prepared dishes set, and evaluate the uniformity of the pre-prepared dishes intention of the users who frequently purchase pre-prepared dishes based on the comprehensive change index of the intention:

[0136] The calculation formula of the intention comprehensive change index is: ;in, is the comprehensive change index of intention, is the number of key features, is the key feature number, It is The weight of the key features, It is The standard deviation calculation results corresponding to the key features.

[0137] The larger the comprehensive change index of intention, the smaller the uniformity of pre-prepared meal intention among users who frequently purchase pre-prepared meals; the smaller the comprehensive change index of intention, the greater the uniformity of pre-prepared meal intention among users who frequently purchase pre-prepared meals. The greater the uniformity, the more consistent the users' purchasing behavior tends to be and their preferences are more concentrated.

[0138] By systematically analyzing the purchase and browsing behavior of users who frequently purchase pre-prepared meals, establishing a set of intended pre-prepared meals, and evaluating the uniformity of intended pre-prepared meals by calculating the intended comprehensive change index, this solution significantly improves the market responsiveness and customer service quality of the e-commerce platform. The intended comprehensive change index enables the platform to identify and predict the dynamic changes in user needs, thereby adjusting the recommendation algorithm and marketing strategy to meet user needs more accurately. This not only enhances user satisfaction and loyalty, but also reduces operating costs and improves operational efficiency by optimizing product supply chain and inventory management.

[0139] The update of search results of users who frequently purchase pre-prepared meals is analyzed to evaluate the divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals. Specifically:

[0140] The process from when users who frequently purchase pre-prepared meals start searching for pre-prepared meals to when they exit the e-commerce platform is marked as a search task. The search task refers to the process in which users who frequently purchase pre-prepared meals actively search for pre-prepared meals on the e-commerce platform.

[0141] A search task set is established, which includes multiple search tasks of users who frequently purchase pre-prepared meals recently.

[0142] It is determined whether there is a re-search situation in the search task, and the search task in the search task set that has the re-search situation is marked as a divergent search task.

[0143] Among them, the re-search situation refers to the situation where users who frequently purchase pre-prepared meals did not find satisfactory results after one search attempt, so they made additional search attempts.

[0144] The ratio of the number of divergent search tasks in the search task set to the total number of search tasks in the search task set is marked as the divergent search ratio.

[0145] Record the total number of searches performed by users who frequently purchase pre-prepared meals in the divergent search task, including the initial search and all re-searches.

[0146] For each search in the divergent search task, including the initial search and all re-searches, the browsing time of users who frequently purchase pre-prepared meals on the search results page is recorded.

[0147] Calculate the ratio of search and browse time, the expression is: ;in, is the ratio of the browsing time of the search again, which indicates the The ratio of the search result page view time of the second search to the search result page view time of the initial search; Indicates The search results page browsing time for the next search; Represents the search results page view time for the initial search.

[0148] If the re-search browsing time ratio is high, it means that users who frequently purchase pre-prepared meals have found more relevant content in subsequent searches, or need more time to evaluate search results, indicating that the search intentions of users who frequently purchase pre-prepared meals have high uncertainty and divergence.

[0149] To reflect the importance or influence of each re-search, Weighting can be based on various factors, such as search sequence number or the intensity of behavior of users who frequently purchase pre-made meals.

[0150] For example, the weights can vary Increases and decreases. In this case, the initial search and subsequent re-searches of users who frequently purchase pre-prepared meals are considered more important, because these are usually the stages in which users who frequently purchase pre-prepared meals try to find content that meets their needs. As the number of re-searches increases, the new information in each re-search may decrease, so re-searches shortly after the initial search may better reflect the dissatisfaction of users who frequently purchase pre-prepared meals with the initial response of the recommendation system, and therefore are given a higher weight; for example, if users who frequently purchase pre-prepared meals conduct multiple searches in a short period of time, it may mean that users who frequently purchase pre-prepared meals are trying to find more relevant content, and these searches may be particularly critical to understanding the real needs of users who frequently purchase pre-prepared meals.

[0151] By assigning weights to the re-search browsing time ratios in the divergent search tasks and summing them up, the search intention divergence value of each divergent search task is calculated, and its expression is: ;

[0152] in, is the search intent divergence value, The weight is determined based on the order of re-search or other factors.

[0153] The larger the search intention divergence value, the more likely it is that users who frequently purchase pre-prepared meals continue to discover new and relevant information in multiple searches, or that they need more time and effort to meet their needs. This suggests that the search intention of users who frequently purchase pre-prepared meals has a strong divergence. It may also indirectly reflect that users who frequently purchase pre-prepared meals are dissatisfied with their current searches, indicating that the e-commerce platform's recommendations are not accurate enough.

[0154] The search intention divergence values ​​of all divergent search tasks in the search task set are summed to obtain the set search intention divergence value.

[0155] The divergent search ratio and the set search intention divergence value are dimensionless, and the divergent search ratio after dimensionless processing and the set search intention divergence value are weighted summed to calculate the search intention divergence index, which is expressed as follows: ;

[0156] in, is the search intention divergence index, are the divergent search ratio and the set search intention divergence value after dimensionless processing, are the weights of the divergent search ratio and the divergent value of the set search intention after dimensionless processing, and Both are greater than 0.

[0157] The search intention divergence index is used to evaluate the degree of divergence in pre-prepared meal intention among users who frequently purchase pre-prepared meals. The larger the search intention divergence index, the greater the degree of divergence in pre-prepared meal intention among users who frequently purchase pre-prepared meals.

[0158] Among them, the degree of divergence in pre-prepared meal intention among users who frequently purchase pre-prepared meals refers to the diversity of demands and variability in search behavior shown by users who frequently purchase pre-prepared meals when searching for pre-prepared meals.

[0159] By deeply analyzing divergent search tasks and calculating the search intention divergence index, e-commerce platforms can more accurately capture and understand user demand changes and satisfaction. This not only helps the platform optimize search algorithms and recommendation strategies, reducing the number of users' re-searches, but also improves user experience, increases user stay time on the platform and purchase conversion rate.

[0160] Based on the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals and the divergence of their pre-prepared meal intention, the urgency of updating the user profile of the users who frequently purchase pre-prepared meals is evaluated, and it is determined whether to update the user profile immediately. Specifically:

[0161] The greater the comprehensive change index of intention, the more urgent it is to update the user profile of users who frequently purchase pre-prepared meals.

[0162] The greater the search intention divergence index, the more urgent it is to update the user profile of users who frequently purchase pre-prepared meals.

[0163] When the comprehensive change index of intention is less than or equal to the comprehensive change threshold of intention, and the search intention divergence index is less than or equal to the search intention divergence threshold, it is determined not to update the user portrait immediately; otherwise, it is determined to update the user portrait immediately.

[0164] When it is determined not to update the user portrait immediately, the user portrait is updated based on a preset update frequency.

[0165] The intention comprehensive change index and search intention divergence index are used as the basis for judging the urgency of updating the user portrait, based on the immediacy requirements of the portrait due to the dynamic changes in user behavior. These two indexes reflect the changes in user needs and preferences: a high intention comprehensive change index indicates that the user's preferences and needs have changed significantly in the considered feature dimensions. This may be caused by changes in the market environment, the launch of new products, or changes in the user's personal situation. It indicates that the user's current portrait is no longer sufficient to accurately reflect their latest consumption tendencies, so it needs to be updated immediately to maintain the relevance and effectiveness of services and recommendations. A high search intention divergence index indicates that the user is more exploratory in the search process and may be looking for new or different products, which usually means that the user's needs are developing or changing. This increase in divergence also indicates that the user portrait needs to be updated to reflect this change.

[0166] When the comprehensive change index of intention and the search intention divergence index are both lower than their respective thresholds, it means that the user's behavior is relatively stable and does not change much. At this time, the portrait can be updated according to the preset update frequency without immediate update. This helps optimize resource utilization and avoid overly frequent data processing and analysis.

[0167] The setting of the comprehensive change threshold of intention is mainly based on the statistical analysis of historical data, aiming to capture major changes in user behavior, thereby triggering the update of the user portrait. First, by collecting the comprehensive change index data of intention of all users over a period of time, its mean and standard deviation are calculated. The threshold is usually set to the mean plus a certain multiple of the standard deviation (such as twice the standard deviation). This method ensures that the update is triggered only when the user's behavior changes significantly deviate from the norm, thereby balancing the reaction sensitivity and operation frequency. In addition, the setting of the threshold also needs to take into account the business cycle and market dynamics. For example, during the launch of a new product or a major marketing campaign, the threshold may need to be temporarily adjusted to adapt to market changes more flexibly. This dynamic adjustment strategy helps the platform maintain sensitivity to changes in user behavior while avoiding overly frequent updates of user portraits, ensuring the effective use of resources.

[0168] The setting of the search intention divergence threshold focuses on evaluating the breadth and depth of users' exploration in their search behavior. Similarly, by analyzing the search intention divergence index in historical data, calculating its mean and standard deviation, the threshold is usually set to the mean plus two to three times the standard deviation to identify users whose search behavior is significantly different from the normal pattern. Setting such a threshold helps to distinguish between daily minor fluctuations and changes in behavioral trends that really need attention, which is crucial for timely adjustment of recommendation systems and content display. In addition, threshold setting also needs to dynamically consider user feedback and satisfaction survey results to ensure that the actual application of the threshold is closely related to user satisfaction and business goals. This setting method enables e-commerce platforms to effectively respond to the challenges of user needs and market changes while maintaining the quality of user experience.

[0169] In addition, the e-commerce platform can monitor the urgency of updating the user profile based on the comprehensive change index of intention and the divergence index of search intention. When these indices exceed the predetermined threshold, it indicates that the user's preferences may be undergoing significant changes, which is a signal for the e-commerce platform to take action. In this case, the platform can work with the pre-prepared meal producers to implement a dynamic pricing strategy. For example, special offers can be provided to consumers whose user profiles show significant changes, aiming to motivate them to try new or improved pre-prepared meal products. This strategy can not only increase the trial rate of new products, but also attract and maintain consumer interest through special promotions.

[0170] By combining the urgency with which the user profile needs to be updated and the user's purchase history, the e-commerce platform can evaluate the user's interest in new products. Based on the user's interest in new products, the e-commerce platform can send a signal to the producer of pre-prepared meals: if the system detects that a user's profile needs to be updated urgently (indicating that their preferences may have changed significantly), the system can provide special discounts.

[0171] Through this strategy, e-commerce platforms can accurately identify and motivate users who are potentially interested in new products, and promote rapid trial and acceptance of new products. This not only helps producers quickly collect market feedback and optimize product design, but also accelerates the market penetration of new products and improves user satisfaction. At the same time, this strategy can also enhance the user's purchasing experience and increase user loyalty and activity on the platform by providing personalized offers.

[0172] By analyzing the entropy of keywords searched by users who frequently purchase pre-prepared meals, the intensity of demand for diversity in the pre-prepared meal recommendations on e-commerce platforms is evaluated. Specifically:

[0173] S201: Collect all search queries for pre-prepared dishes conducted by users who frequently purchase pre-prepared dishes on the e-commerce platform within a preset period. The search queries include keywords entered by users who frequently purchase pre-prepared dishes, search frequency, and search date:

[0174] Collect all the search queries of regular purchasers of pre-made dishes for pre-made dishes within a preset range to analyze the diversity of their search behaviors.

[0175] Regular purchaser ID of pre-made dishes: Identifies each unique regular purchaser of pre-made dishes.

[0176] Keywords: The words entered by regular purchasers of pre-made dishes when searching for pre-made dishes.

[0177] Search frequency: The number of times a keyword is searched within a specific time period.

[0178] Search date: The specific date when the search is conducted.

[0179] S202: Extract all independent keywords. Organize and clean the keywords to remove meaningless filler words:

[0180] Extract all the keywords entered by regular purchasers of pre-made dishes from the search data. Use text processing techniques to remove meaningless filler words (such as "de", "he", etc.) and eliminate noisy data (such as typos or irrelevant words).

[0181] Filler words: Words that do not carry specific meanings in the text and are not substantially helpful for analysis.

[0182] Noisy data: Incorrect inputs or keywords irrelevant to pre-made dishes.

[0183] S203: Count the number of times each keyword appears in all search queries, and then calculate the occurrence frequency of each keyword:

[0184] For each keyword, count the total number of times it appears in the entire dataset.

[0185] The formula for calculating the occurrence frequency of each keyword is:

[0186] ; where is the occurrence frequency of the th keyword, is the th keyword's occurrence count, is the total occurrence count of all keywords.

[0187] S204: Apply the information entropy formula to calculate the search diversity entropy evaluation index:

[0188] Use the information entropy formula to calculate the search diversity entropy evaluation index to quantify the diversity of the search behaviors of regular purchasers of pre-made dishes. Its expression is: ; is the search diversity entropy evaluation index, is the total number of different keywords.

[0189] The larger the search diversity entropy evaluation index, the higher the diversity demand intensity of the pre-prepared meal recommendations on the e-commerce platform, indicating that the search behavior of users who frequently purchase pre-prepared meals is highly diverse. The recommendation system should increase the diversity of pre-prepared meals in the recommendation list and display a wider range of pre-prepared meal types. The smaller the search diversity entropy evaluation index, the more concentrated the search of users who frequently purchase pre-prepared meals. The recommendation system should reduce the diversity of recommendations and recommend more specific types of pre-prepared meals that users who frequently purchase pre-prepared meals may be interested in.

[0190] in, is the total number of different keywords, that is, the number of unique keyword types in the dataset. It represents the total number of unique keyword types that appear in the search process of users who often purchase pre-prepared meals. It is the total number of occurrences of all keywords, that is, the cumulative number of times a keyword appears in all searches. Regardless of whether the keyword is repeated or not, every occurrence is counted. . The focus is on keyword diversity (how many different keywords were searched), while What we focus on is the frequency of keyword usage (the total number of times all keywords are searched).

[0191] Through a comprehensive analysis of the homogeneity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the divergence of their pre-prepared meal intentions, and the intensity of the diversity demand for pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms to users who frequently purchase pre-prepared meals is adjusted, specifically:

[0192] Since the larger the intention comprehensive change index is, the smaller the uniformity of the pre-prepared food intention of users who often purchase pre-prepared food is; the larger the search intention divergence index is, the greater the divergence of the pre-prepared food intention of users who often purchase pre-prepared food is; the larger the search diversity entropy evaluation index is, the higher the diversity demand intensity of the pre-prepared food recommendation of the e-commerce platform is. It can be seen that the intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are all proportional to the category breadth of the pre-prepared food recommended by the e-commerce platform to users who often purchase pre-prepared food, that is, the larger the intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are, the greater the category breadth of the pre-prepared food recommended by the e-commerce platform to users who often purchase pre-prepared food should be adjusted.

[0193] The intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are normalized, and after assigning preset proportional coefficients to the normalized intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index, the recommended extensive evaluation coefficient is calculated.

[0194] The specific implementation method of calculating the recommended broad evaluation coefficient is not specifically limited here. Any method that can calculate the recommended broad evaluation coefficient by assigning the normalized intention comprehensive change index, search intention divergence index, and search diversity entropy evaluation index to preset proportional coefficients is acceptable. In order to implement the technical solution of the present invention, the present invention provides a specific method for calculating the recommended broad evaluation coefficient, and its expression is: ;

[0195] in, To recommend a broad assessment coefficient, is the base of natural logarithm, are the preset proportional coefficients of the intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index, respectively, and Both are greater than 0.

[0196] The relationship between the recommendation breadth evaluation coefficient and the breadth of the categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals is: the larger the recommendation breadth evaluation coefficient, the greater the breadth of the categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals should be adjusted; that is, the larger the recommendation breadth evaluation coefficient, the more the breadth of the recommended categories should be increased to display more diverse pre-prepared meal options to users.

[0197] This regulation can be achieved through linear or nonlinear models. For example, a threshold can be set. When the recommendation extensive evaluation coefficient exceeds a certain threshold, the recommendation diversity is greatly increased.

[0198] Alternatively, thresholds are set to determine different thresholds of the recommendation extensive evaluation coefficient to distinguish the levels of recommendation extensiveness, for example, low, medium, and high levels, corresponding to the low, medium, and high value intervals of the recommendation extensive evaluation coefficient.

[0199] Suppose we want the recommendation system to be more sensitive to increase the diversity of recommendations when the recommendation broad evaluation coefficient is high, and to maintain a more stable diversity of recommendations when the recommendation broad evaluation coefficient is low. We can use an exponential function or a logarithmic function to implement this nonlinear relationship. Use an exponential function to adjust the breadth of recommended categories: ;in, is the adjusted recommended diversity level; Both are adjustment coefficients used to control the growth rate and baseline diversity level of the function.

[0200] is the growth rate control parameter that determines the rate at which the diversity of recommendations increases when the recommendation broad evaluation coefficient increases. The value of will make the diversity more sensitive to the recommended broad evaluation coefficient, that is, small changes in the recommended broad evaluation coefficient will lead to large changes in the recommended diversity.

[0201] is the baseline diversity level, which is the benchmark value of the recommended diversity level. , different initial recommendation diversity levels can be set.

[0202] Among them, the recommendation diversity level is a numerical expression of the breadth of categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals.

[0203] By introducing a recommendation extensive evaluation coefficient, the recommendation of pre-prepared meals can dynamically adapt to the diverse needs of users. This approach not only enhances the flexibility of the recommendation system, but also improves its sensitivity, enabling it to respond instantly to subtle changes in user preferences. Through nonlinear adjustments, such as exponential functions, the system can quickly expand the scope of recommendations when the user's willingness to explore increases, thereby greatly improving user satisfaction and engagement. In addition, this solution also helps to increase the exposure and trial rate of new products, ultimately driving sales growth and user loyalty. This data-driven recommendation extensiveness adjustment strategy brings innovative user experience optimization methods to e-commerce platforms.

[0204] Example 2

[0205] The difference between Example 2 of the present invention and Example 1 is that this example introduces a pre-prepared meal recommendation system.

[0206] Figure 2 A structural schematic diagram of a prepared dish recommendation system of the present invention is given, which includes a user segmentation module, an intention uniformity evaluation module, an intention divergence evaluation module, a portrait update judgment module, a recommendation diversity evaluation module and a recommendation extensive adjustment module.

[0207] User segmentation module: obtain potential recommended users of pre-prepared dishes based on the database, and divide the potential recommended users into users who frequently purchase pre-prepared dishes and users who do not frequently purchase pre-prepared dishes.

[0208] Intention uniformity evaluation module: Analyze the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes, screen out the intended pre-prepared dish set, analyze the degree of change in the characteristics of the intended pre-prepared dish set, and evaluate the uniformity of pre-prepared dish intention of users who frequently purchase pre-prepared dishes.

[0209] Intention divergence evaluation module: Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals.

[0210] Portrait update judgment module: Based on the uniformity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals and the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the urgency of updating the user portraits of users who frequently purchase pre-prepared meals is evaluated, and a judgment is made as to whether the user portraits should be updated immediately.

[0211] Recommendation diversity evaluation module: By analyzing the entropy of search keywords of users who frequently purchase pre-prepared meals, the diversity demand intensity of pre-prepared meal recommendations on e-commerce platforms is evaluated.

[0212] Recommendation breadth adjustment module: Through a comprehensive analysis of the uniformity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, and the intensity of demand for diversity of pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms is adjusted.

[0213] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0214] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0215] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0216] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0217] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0218] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0219] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0220] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0221] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0222] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for recommending pre-prepared dishes, characterized in that: The steps include: Obtain potential recommended users of pre-prepared dishes based on the database, and divide the potential recommended users into users who frequently purchase pre-prepared dishes and users who do not frequently purchase pre-prepared dishes; Analyze the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes, screen out the intended pre-prepared dish set, analyze the degree of change in the characteristics of the intended pre-prepared dish set, and evaluate the uniformity of the intended pre-prepared dish of users who frequently purchase pre-prepared dishes; Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals; Based on the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals and the divergence of their pre-prepared meal intention, the urgency of updating the user profile of the users who frequently purchase pre-prepared meals is evaluated, and it is determined whether to update the user profile immediately. By analyzing the entropy of keywords searched by users who frequently purchase prepared meals, the intensity of demand for diversity in the recommendations of prepared meals on e-commerce platforms is evaluated; Through a comprehensive analysis of the homogeneity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the divergence of their pre-prepared meal intentions, and the intensity of the diversity of pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms to users who frequently purchase pre-prepared meals is adjusted; The intended pre-prepared dishes of users who frequently purchase pre-prepared dishes are analyzed, and a set of intended pre-prepared dishes is obtained by screening. The degree of change in the characteristics of the set of intended pre-prepared dishes is analyzed, and the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes is evaluated. Specifically: S101: establishing a desired pre-prepared meal set by identifying frequently purchased and browsed pre-prepared meal products based on historical purchase data and browsing behavior of users who frequently purchase pre-prepared meals; S102: Collecting transaction records, user reviews, browsing data and other interactive data of the intended pre-prepared meal set; S103: extracting key features for each intended pre-prepared dish in the intended pre-prepared dish set, where the key features include convenience, category, and homogeneity; S104: Calculate the standard deviation of each key feature, and quantify the volatility of the key feature change according to the standard deviation calculation result of the key feature; S105: assigning a weight to each key feature, performing weighted processing on the standard deviation of the key features and then summing them up, obtaining a comprehensive change index of intention for the intended pre-prepared meal set, and evaluating the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals based on the comprehensive change index of intention; The calculation formula of the intention comprehensive change index is: Among them, CI is the comprehensive change index of intention, k is the number of key features, j is the number of key features, uj is the weight of the jth key feature, and σj is the standard deviation calculation result corresponding to the jth key feature; The update of search results of users who frequently purchase pre-prepared meals is analyzed to evaluate the divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals. Specifically: The process from when users who frequently purchase pre-prepared food start searching for pre-prepared food to when they exit the e-commerce platform is marked as a search task; Establish a search task set, which includes multiple search tasks of users who frequently purchase pre-prepared dishes recently; Determine whether there is a re-search situation in the search task, and mark the search task with the re-search situation in the search task set as a divergent search task; The ratio of the number of divergent search tasks in the search task set to the total number of search tasks in the search task set is marked as the divergent search ratio; Record the total number of searches conducted by users who frequently purchase pre-prepared meals in the divergent search task; For each search in the divergent search task, record the browsing time of users who frequently purchase pre-prepared meals on the search results page; Calculate the re-search browsing time ratio; The search intention divergence value of each divergent search task is calculated by assigning weights to the re-search browsing time ratios in the divergent search task and summing them up; Sum the search intention divergence values ​​of all divergent search tasks in the search task set to obtain the set search intention divergence value; The divergent search ratio and the set search intention divergence value are dimensionless, and the divergent search ratio after dimensionless processing and the set search intention divergence value are weighted and summed to calculate the search intention divergence degree index, whose expression is: Sf = α*Fs+β*Js; where Sf is the search intention divergence degree index, Fs and Js are the divergent search ratio after dimensionless processing and the set search intention divergence value, α and β are the weights of the divergent search ratio after dimensionless processing and the set search intention divergence value, and α and β are both greater than 0; By analyzing the entropy of keywords searched by users who frequently purchase pre-prepared meals, the intensity of demand for diversity in the pre-prepared meal recommendations on e-commerce platforms is evaluated. Specifically: S201: Collect all search queries for pre-prepared dishes conducted by users who frequently purchase pre-prepared dishes on the e-commerce platform within a preset period; the search queries include keywords input by users who frequently purchase pre-prepared dishes, search frequencies, and search dates; S202: extract all independent keywords; sort and clean the keywords and remove meaningless filler words; S203: Count the number of times each keyword appears in all search queries, and then calculate the frequency of occurrence of each keyword; the frequency of occurrence calculation formula of each keyword is: Among them, pv is the frequency of occurrence of the vth keyword, bv is the number of occurrences of the vth keyword, and N is the total number of occurrences of all keywords; S204: Apply the information entropy formula to calculate the search diversity entropy evaluation index, which is expressed as: Hf is the search diversity entropy evaluation index, and f is the total number of different keywords; Through a comprehensive analysis of the homogeneity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the divergence of their pre-prepared meal intentions, and the intensity of the diversity demand for pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms to users who frequently purchase pre-prepared meals is adjusted, specifically: The intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are normalized, and the normalized intention comprehensive change index, search intention divergence index and search diversity entropy evaluation index are respectively assigned preset proportional coefficients to calculate the recommendation extensive evaluation coefficient, which is expressed as follows: Among them, TP is the recommendation extensive evaluation coefficient, e is the natural logarithm base, θ1, θ2, and θ3 are the preset proportional coefficients of the intention comprehensive change index, the search intention divergence index, and the search diversity entropy evaluation index, respectively, and θ1, θ2, and θ3 are all greater than 0; The relationship between the recommendation breadth evaluation coefficient and the breadth of categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals is: the larger the recommendation breadth evaluation coefficient, the greater the breadth of categories of pre-prepared meals recommended by the e-commerce platform to users who frequently purchase pre-prepared meals should be adjusted.

2. A method for recommending prepared dishes according to claim 1, characterized in that: Based on the database, potential recommended users of pre-prepared dishes are obtained, and the potential recommended users are divided into users who often purchase pre-prepared dishes and users who do not often purchase pre-prepared dishes. Specifically: The data collected in the database includes basic user information, transaction information, browsing history, and product information; Construct features for user classification based on the data collected from the database. The features used for user classification include purchase frequency, purchase preference, consumption level, and activity level; Label users according to purchase frequency, divide the data into training and test sets, and use the random forest algorithm to train the model; The trained model is used for user classification, and potential recommendation users are divided into users who frequently purchase pre-prepared meals and users who do not frequently purchase pre-prepared meals.

3. A method for recommending prepared dishes according to claim 2, characterized in that: Based on the uniformity of the pre-prepared meal intention of the users who frequently purchase pre-prepared meals and the divergence of their pre-prepared meal intention, the urgency of updating the user profile of the users who frequently purchase pre-prepared meals is evaluated, and it is determined whether to update the user profile immediately. Specifically: When the comprehensive change index of intention is less than or equal to the comprehensive change threshold of intention, and the search intention divergence index is less than or equal to the search intention divergence threshold, it is determined that the user profile will not be updated immediately; Otherwise, it is determined that the user profile should be updated immediately; When it is determined that the user portrait is not to be updated immediately, the user portrait is updated based on a preset update frequency.

4. A pre-prepared food recommendation system, used to implement a pre-prepared food recommendation method according to any one of claims 1 to 3, characterized in that: It includes user segmentation module, intention uniformity assessment module, intention divergence assessment module, portrait update judgment module, recommendation diversity assessment module and recommendation extensive adjustment module; User segmentation module: obtains potential recommended users of pre-prepared dishes based on the database, and divides the potential recommended users into users who frequently purchase pre-prepared dishes and users who rarely purchase pre-prepared dishes; Intention uniformity evaluation module: Analyze the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes, screen out the intended pre-prepared dishes set, analyze the degree of change in the characteristics of the intended pre-prepared dishes set, and evaluate the uniformity of the intended pre-prepared dishes of users who frequently purchase pre-prepared dishes; Intention divergence evaluation module: Analyze the re-update of search results of users who frequently purchase pre-prepared meals, and evaluate the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals; Profile update judgment module: Based on the uniformity of the pre-prepared food intention of the pre-prepared food frequent purchasing users and the divergence of the pre-prepared food intention of the pre-prepared food frequent purchasing users, the urgency of updating the user profile of the pre-prepared food frequent purchasing users is evaluated, and it is judged whether to update the user profile immediately; Recommendation diversity evaluation module: By analyzing the entropy of search keywords of users who frequently purchase pre-prepared dishes, the diversity demand intensity of pre-prepared dishes recommended by e-commerce platforms is evaluated; Recommendation breadth adjustment module: Through a comprehensive analysis of the uniformity of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, the degree of divergence of pre-prepared meal intentions of users who frequently purchase pre-prepared meals, and the intensity of demand for diversity of pre-prepared meal recommendations on e-commerce platforms, the breadth of categories of pre-prepared meals recommended by e-commerce platforms is adjusted.