Product recommendation method and system based on hierarchical information

Abstract

The invention relates to the field of product recommendation, and provides a product recommendation method and system based on hierarchical information in order to improve the product recommendation accuracy, the information of a user and the use preference, use evaluation and use scene of a product are obtained through hierarchical data collection, association rules among the hierarchical information are mined based on an Apriori algorithm, and the product recommendation accuracy is improved. Dividing user groups by adopting a K-means clustering algorithm based on association rules and hierarchical information, and generating a Top-N recommendation list according to purchase records corresponding to the user groups to which the to-be-recommended users belong; and a'demand-behavior-feedback 'closed loop is formed through hierarchical data acquisition, and accurate recommendation is driven, so that the product recommendation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of product recommendation, and in particular to a product recommendation method and system based on hierarchical information. Background Art

[0002] In the food and beverage industry, especially in the baijiu (white spirits) sector, quantitative assessment of user experience is crucial for product development and market adaptation. The level of comfort experienced by drinkers during consumption directly influences their overall evaluation of baijiu and their willingness to consume again. Comfort encompasses not only the taste and aroma of the liquor but also the physical and psychological reactions after drinking. These factors collectively determine whether consumers are willing to continue choosing a particular brand or type of baijiu.

[0003] Traditional liquor recommendation methods rely on manual experience or simple rules (such as price range matching), and lack in-depth analysis of user taste preferences, drinking scenarios and physiological reactions, resulting in recommendation results being out of touch with real needs. Summary of the Invention

[0004] In order to improve the accuracy of product recommendations, this application provides a product recommendation method and system based on hierarchical information.

[0005] The technical solution adopted by the present invention to solve the above problems is: Product recommendation methods based on hierarchical information include: Step 1: Conduct hierarchical information collection, including basic information layer, behavior layer, experience layer and scenario layer. The basic information layer is used to collect user information; the behavior layer is used to collect user preferences for product usage; the experience layer is used to evaluate user experience with the product; and the scenario layer is used to collect product usage scenarios. Step 2: Use the Apriori algorithm to mine the association rules between each layer of information, and use the K-means clustering algorithm to divide user groups based on the association rules and layer information; Step 3: Obtain the hierarchical information of the user to be recommended, obtain the matching degree between the user to be recommended and the user group based on cosine similarity, and generate a Top-N recommendation list based on the purchase records of the corresponding user group.

[0006] Furthermore, the recommended products are alcoholic products.

[0007] Furthermore, the basic information layer includes: gender, age, weight and personal income; The behavioral layer includes: drinking frequency, drinking amount, alcohol content, consumption price and taste preference; The experience layer includes: positive experience and negative experience; The scenario layer includes: business banquets, family gatherings, celebrations, social occasions and drinking alone.

[0008] Furthermore, step 1 collects information through a questionnaire.

[0009] Furthermore, step 1 also includes: pre-processing the collected information, including: eliminating invalid questionnaires, filling in missing values ​​and data standardization.

[0010] Furthermore, invalid questionnaires include: age less than 18 years old, drinking frequency of 0, and questionnaire filling time lower than the threshold.

[0011] Furthermore, in step 3, the purchase record includes the purchase quantity and purchase frequency of the product. The specific method of generating a Top-N recommendation list based on the purchase record is: setting weights for the purchase quantity and purchase frequency respectively; performing weighted summation of the purchase quantity and purchase frequency to obtain a comprehensive score for the product; and generating a Top-N recommendation list based on the comprehensive score.

[0012] Furthermore, when using the Apriori algorithm to mine association rules between hierarchical information, the parameters are set as support ≥ 0.3 and confidence ≥ 0.7.

[0013] Product recommendation system based on hierarchical information, including: The layered information collection module is used to collect layered information, including the basic information layer, behavior layer, experience layer and scenario layer; User group segmentation module: Use the Apriori algorithm to mine the association rules between each layer of information, and use the K-means clustering algorithm to divide user groups based on the association rules and layer information; Recommendation list generation module: obtains the hierarchical information of the user to be recommended, obtains the matching degree between the user to be recommended and the user group based on cosine similarity, and generates a Top-N recommendation list based on the purchase records of the corresponding user group.

[0014] Furthermore, the hierarchical information collection module collects hierarchical information through a questionnaire.

[0015] Compared with the existing technology, the present invention has the following beneficial effects: obtaining user's own information, product usage preferences, usage evaluations and usage scenarios through hierarchical data collection, mining the association rules between each hierarchical information based on the Apriori algorithm, dividing user groups based on the association rules and hierarchical information using the K-means clustering algorithm, and generating a Top-N recommendation list according to the purchase records corresponding to the user group to which the recommended user belongs; forming a "demand-behavior-feedback" closed loop through hierarchical data collection to drive accurate recommendations, thereby improving the accuracy of product recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Flowchart of a method for recommending products based on hierarchical information; Figure 2 This is the data preprocessing flow chart; Figure 3 Design a flow chart for a stratified questionnaire. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] Take liquor recommendation as an example. Figure 1 As shown in FIG, the product recommendation method based on hierarchical information includes: Step 1: Conduct hierarchical information collection, including the basic information layer, behavior layer, experience layer, and scenario layer. The basic information layer defines the user's physiological and economic boundaries and is used to collect user information, such as gender, age, weight, personal income, and other demographic data. The behavior layer quantifies product interaction preferences and is used to collect user preferences for product usage, covering drinking frequency, single-serving amount, consumption price, alcohol content preference, and taste preference options. Each taste option is rated on a 1-10 scale. The experience layer captures physiological and psychological feedback and is used to evaluate the user's experience with the product, including both negative and positive experiences. Negative experiences include thirst, loss of appetite, nausea, vomiting, fatigue, anxiety, depression, difficulty concentrating, dizziness, palpitations, headaches, and difficulty sleeping. Positive experiences include pleasure, relaxation, stress relief, excitement, increased appetite, improved sleep, confidence, comfort, intoxication, happiness, satisfaction, and warmth. The scenario layer maps usage scenarios and is used to collect product usage scenarios, including business dinners, family gatherings, celebrations, social occasions, and drinking alone. The four layers work together to form a closed loop of "demand-behavior-feedback" to drive accurate recommendations. To facilitate the acquisition of hierarchical information, this embodiment uses a questionnaire to obtain it.

[0019] In order to improve the accuracy of recommendations, the collected information needs to be preprocessed, including eliminating invalid questionnaires, filling in missing values ​​and data standardization, such as Figure 2 Invalid questionnaires include those with an age of less than 18, a drinking frequency of 0, and a questionnaire filling time less than a threshold (e.g., 30 seconds). To improve data processing efficiency, this embodiment performs data verification (whether the age is less than 18, whether the drinking frequency is 0, and whether the questionnaire filling time is less than a threshold) when the questionnaire is submitted. If any of the verifications fails, the questionnaire will be directly discarded. Figure 3 Missing values ​​are filled using interpolation. For example, for missing taste ratings, the mean value of similar users (e.g., the same age or gender group) is used for interpolation. Data normalization uses the Z-score method.

[0020] Step 2: Use the Apriori algorithm to discover association rules between each layer of information. Based on these association rules and the layered information, use the K-means clustering algorithm to segment user groups. Using the four-dimensional structured data collected through the stratified questionnaire, convert categorical data (such as scenario labels) into Boolean variables, and bin numerical data (such as taste ratings) (e.g., 8-10 is considered a "high rating"). Use the Apriori algorithm to traverse all possible item set combinations and select strong association rules that meet support ≥ 0.3 and confidence ≥ 0.7.

[0021] According to the association rules, the clustering feature weights are dynamically adjusted to improve the classification accuracy. The feature weight calculation formula is: w i = basic weight × (1 + rule confidence × rule support), where w i is the feature weight, and the rule confidence and rule support come from Apriori output.

[0022] The input feature vector includes the standardized values ​​of the four-layer data and the corresponding weights. The optimal number of clusters is determined by the elbow rule. User samples that meet the strong association rules are preferentially selected as the initial cluster centers. The intra-cluster distance is iteratively optimized until convergence.

[0023] Step 3: Obtain the hierarchical information of the user to be recommended, determine the match between the user to be recommended and the user group based on cosine similarity, and generate a Top-N recommendation list based on the purchase history of the corresponding user group. Input the feature vectors of the user to be recommended and the grouped users (containing four layers of data), calculate cosine similarity, and select the Top-N users with the highest similarity as neighbors. Generate a Top-N recommendation list based on the purchase history of neighboring users. Purchase history includes the quantity and frequency of purchases of a product. The specific method for generating a Top-N recommendation list based on purchase history is as follows: set weights for the purchase quantity and frequency of purchases respectively; the specific weight values ​​can be set according to actual needs; perform a weighted sum of the purchase quantity and frequency of purchases to obtain a comprehensive score for the product; and generate a Top-N recommendation list based on the comprehensive score.

[0024] In addition, this embodiment also provides a product recommendation system based on hierarchical information, including: Layered information collection module, used to collect layered information, including basic information layer, behavior layer, experience layer and scenario layer; User group segmentation module: Use the Apriori algorithm to mine the association rules between each layer of information, and use the K-means clustering algorithm to divide user groups based on the association rules and layer information; Recommendation list generation module: obtains the hierarchical information of the user to be recommended, obtains the matching degree between the user to be recommended and the user group based on cosine similarity, and generates a Top-N recommendation list based on the purchase records of the corresponding user group.

[0025] Among them, the hierarchical information collection module collects hierarchical information through questionnaires.

Claims

1. A product recommendation method based on hierarchical information, characterized in that: include: Step 1: Conduct hierarchical information collection, including basic information layer, behavior layer, experience layer and scenario layer. The basic information layer is used to collect user information; the behavior layer is used to collect user preferences for product usage; the experience layer is used to evaluate user experience with the product; and the scenario layer is used to collect product usage scenarios. Step 2: Use the Apriori algorithm to mine the association rules between each layer of information, and use the K-means clustering algorithm to divide user groups based on the association rules and layer information; Step 3: Obtain the hierarchical information of the user to be recommended, obtain the matching degree between the user to be recommended and the user group based on cosine similarity, and generate a Top-N recommendation list based on the purchase records of the corresponding user group.

2. The product recommendation method based on hierarchical information according to claim 1, characterized in that: The recommended products are alcoholic products.

3. The product recommendation method based on hierarchical information according to claim 2, characterized in that: The basic information layer includes: gender, age, weight and personal income; The behavioral layer includes: drinking frequency, drinking amount, alcohol content, consumption price and taste preference; The experience layer includes: positive experience and negative experience; The scenario layer includes: business banquets, family gatherings, celebrations, social occasions and drinking alone.

4. The product recommendation method based on hierarchical information according to claim 3, characterized in that: Step 1: Collect information through questionnaires.

5. The product recommendation method based on hierarchical information according to claim 4, characterized in that: Step 1 also includes: pre-processing the collected information, including: eliminating invalid questionnaires, filling in missing values ​​and data standardization.

6. The product recommendation method based on hierarchical information according to claim 5, characterized in that: Invalid questionnaires included: age less than 18 years old, drinking frequency of 0, and questionnaire completion time below the threshold.

7. The product recommendation method based on hierarchical information according to claim 1, characterized in that: In step 3, the purchase record includes the purchase quantity and purchase frequency of the product. The specific method of generating the Top-N recommendation list based on the purchase record is: set the weights of the purchase quantity and purchase frequency respectively; perform weighted summation of the purchase quantity and purchase frequency to obtain the comprehensive score of the product; and generate the Top-N recommendation list based on the comprehensive score.

8. The product recommendation method based on hierarchical information according to any one of claims 1 to 7, characterized in that: When using the Apriori algorithm to mine association rules between hierarchical information, the parameters are set as support ≥ 0.3 and confidence ≥ 0.

7.

9. A product recommendation system based on hierarchical information, characterized in that: include: Layered information collection module, used to collect layered information, including basic information layer, behavior layer, experience layer and scenario layer; User group segmentation module: Use the Apriori algorithm to mine the association rules between each layer of information, and use the K-means clustering algorithm to divide user groups based on the association rules and layer information; Recommendation list generation module: obtains the hierarchical information of the user to be recommended, obtains the matching degree between the user to be recommended and the user group based on cosine similarity, and generates a Top-N recommendation list based on the purchase records of the corresponding user group.

10. The product recommendation system based on hierarchical information according to claim 9, characterized in that: The stratified information collection module collects stratified information through questionnaires.

Images