Beer production personalized customization management system and method based on artificial intelligence
By collecting and analyzing user needs and historical beer recipe data through an artificial intelligence system, personalized beer recipes are automatically generated and optimized, solving the problem that traditional beer production cannot meet personalized customization needs and achieving efficient and low-cost beer production.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- QINGDAO KANGYI BIOTECHNOLOGY CO LTD
- Filing Date
- 2026-02-10
- Publication Date
- 2026-06-05
AI Technical Summary
Traditional beer production methods struggle to meet personalized customization needs, failing to automatically select and adjust suitable recipes based on user requirements. This results in long R&D cycles and high costs, and a lack of in-depth analysis of historical recipe data, hindering the effective discovery of potential correlations between recipes.
The AI-based personalized beer production management system collects user demand information through an online platform, performs hierarchical clustering and association rule mining, analyzes user profiles and historical recipe data, generates personalized beer recipes, and optimizes and adjusts them based on user feedback.
It enables automated and precise beer recipe design, reduces manual intervention, improves efficiency and accuracy, lowers R&D costs, and allows for real-time response to changes in market and user needs, continuously improving products.
Smart Images
Figure CN122155188A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data management technology, specifically to an artificial intelligence-based personalized customization management system and method for beer production. Background Technology
[0002] As one of the world's most popular alcoholic beverages, beer has undergone a long period of development and innovation. However, as consumers' demands for beer taste and quality continue to rise, the beer market is gradually showing a trend towards diversification. Consumers are paying increasing attention to the taste, flavor, and personalization of beer, and traditional beer production methods have many limitations when facing the demand for personalized customization. In particular, in terms of recipe innovation, production flexibility, and responsiveness to consumer customization needs, traditional beer production systems have failed to fully adapt to the rapid changes in the market.
[0003] While some producers have begun to explore more flexible recipe design in recent years to enhance the diversity and innovation of their beer products, this process often relies on manual experience and cumbersome experimental verification, resulting in low efficiency and high costs. For example, for customers requiring personalized beers, current technology cannot find the historical beer recipes that best match their individual needs based on their specific requirements, nor can it provide modification suggestions for those recipes. This necessitates manual selection and adjustment of recipes, which not only increases the development cycle but also wastes resources and incurs unnecessary costs. Furthermore, current technology lacks in-depth analysis of historical recipe data, failing to effectively uncover potential correlations and patterns between recipes. This makes it difficult for producers to extract effective recipe knowledge that can guide future production when faced with a large number of historical recipes. Even with a recipe database, how to automatically select the most suitable recipe from these historical recipes and make targeted adjustments based on customers' personalized needs remains a pressing problem. Summary of the Invention
[0004] The purpose of this invention is to provide an artificial intelligence-based personalized beer production management system and method to solve the problems mentioned in the background art.
[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A personalized beer production management method based on artificial intelligence includes the following steps: Step S100. Collect users' personalized beer demand information through an online platform, extract corresponding demand features from the personalized beer demand information; classify users according to demand features, and construct user profiles based on the classification results, thereby generating a personalized demand dataset; Step S200. Based on the personalized demand dataset, collect the beer purchase records of all users in the same category, search the recipe database based on the beer purchase records, obtain historical recipe data, and analyze the correlation between different historical recipe data; based on the correlation analysis results, obtain the target historical recipe data for the corresponding category of users; Step S300. Collect evaluation information of the same category of users on the batches of beer corresponding to the target historical recipe data through the online platform, and analyze the raw material ratio in the target historical recipe data in combination with the evaluation information to obtain suggestions for modifying the raw material ratio of the target historical recipe data. Step S400: Output the suggested modifications to the raw material ratios in the target historical recipe data to relevant personnel, who will then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, produce personalized beer, collect feedback data from the corresponding user category on the personalized beer, and optimize and adjust the personalized beer recipe based on the feedback data.
[0006] Furthermore, step S100 includes: S101. Collect users' personalized beer demand information through an online platform, wherein the online platform refers to a platform that collects users' personalized beer demand information and users' beer evaluation information; wherein the collection of users' personalized beer demand information is carried out through a questionnaire survey, thereby obtaining a quantitative representation of personalized beer demand information; extract the corresponding demand features from the personalized beer demand information and perform standardization processing to form a demand feature vector Xi, and Xi=[xi1,xi2,...,xin], where Xi represents the vector corresponding to the personalized beer demand information of the i-th user, i represents the user's data number; xi1 represents the first feature value in the personalized beer demand information of the i-th user, xi2 represents the second feature value in the personalized beer demand information of the i-th user, and so on, xin represents the nth feature value in the personalized beer demand information of the i-th user; n represents the number of feature values in the personalized beer demand information; S102. Based on the user's demand feature vector Xi, hierarchical clustering is used to classify the user's demand feature vector, thereby dividing the user into several categories; for each category, the mean value of the corresponding user on each feature dimension is calculated, thereby obtaining the demand feature mean vector Xa_c of the corresponding category, where c represents the category number; based on the demand feature mean vector Xa_c, user profile labels for the corresponding category are generated; based on each user's demand feature vector and user profile label, a personalized demand dataset G is constructed, and G={g1,g2,...,gm}, where g1 represents the demand feature vector and user profile label of the first user, g2 represents the demand feature vector and user profile label of the second user, and so on, gm represents the demand feature vector and user profile label of the m-th user, where m represents the user's data number.
[0007] Furthermore, hierarchical clustering is used to classify the user's demand feature vectors, as detailed below: Calculate the distance d(Xi,Xj) between the demand feature vectors of each pair of users, and the specific calculation formula is as follows: d(Xi,Xj)={∑k∈[1,n],(xik-xjk) 2}^(1 / 2); Where Xi represents the demand feature vector of the i-th user, Xj represents the demand feature vector of the j-th user, xik represents the k-th demand feature value of the i-th user, and xjk represents the k-th demand feature value of the j-th user; k represents the dimension number of the demand feature vector, ranging from 1 to n; Initially, each user's demand feature vector Xi is set as an independent cluster, and there are m clusters in the initial stage. For each user, the two clusters with the smallest distance d(Xi,Xj) are selected and merged to form a new cluster, and the demand feature vector of the new cluster is the weighted average of all demand feature vectors of the two clusters. After merging the clusters, the distance between the new cluster and other clusters is recalculated. The aggregation process is repeated until the predetermined number of clusters is reached.
[0008] Furthermore, step S200 includes: S201. Based on the personalized needs dataset, collect beer purchase records of all users in the same category; for each user in the corresponding category, calculate the purchase frequency of different beers based on the beer purchase records, arrange the purchase frequencies of different beers in descending order, and record the beer number with the highest purchase frequency as the pre-selection number; summarize the pre-selection numbers of all users in the corresponding category, search in the recipe database according to the pre-selection number, find the historical recipe data corresponding to the pre-selection number, and construct a historical recipe dataset; extract raw material data from the historical recipe data and construct a raw material dataset R, where R={r1,r2,...,rs], where r 1 represents the raw material data of the first type of historical formula data, r2 represents the raw material data of the second type of historical formula data, and so on, rs represents the raw material data of the s-th type of historical formula data; and each type of raw material data re represents a vector of its component characteristics, re=[ce1,ce2,...,cef], where ce1 represents the concentration of the first component in the raw material data of the e-th type of historical formula data, ce2 represents the concentration of the second component in the raw material data of the e-th type of historical formula data, and so on, cef represents the concentration of the f-th component in the raw material data of the e-th type of historical formula data, and f represents the number of components in the raw material; S202. Based on the raw material dataset, calculate the similarity of components among different historical formula data to obtain a similarity matrix S between different historical formulas. Here, Suv in the similarity matrix represents the similarity between raw material data ru and rv. Filter historical formula data in the similarity matrix that are greater than or equal to the similarity threshold S0. For the filtered historical formula data, use association rule mining to analyze the correlation between the corresponding historical formula data. The specific analysis process is as follows: Obtain user profile tags for the corresponding categories, and derive the corresponding ingredient combinations based on these tags. Calculate the support, confidence, and lift for different ingredient combinations in the historical recipe data. The formula for calculating support is: Support(A) = |A| / |D|, where |A| represents the frequency in the historical recipe dataset, and |D| represents the total number of recipes in the historical recipe dataset. The formula for calculating confidence is: Confidence(A→B)=Support(A∪B) / Support(A); Wherein, Confidence(A→B) represents the probability of component combination B occurring given that component combination A has occurred; Support(A∪B) represents the frequency with which component combinations A and B co-occur in the historical formula dataset; the formula for calculating the lift is: Lift(A→B)=Support(A∪B) / [Support(A)·Support(B)], where Lift(A→B) represents the ratio of the probability of component combination B occurring given that component combination A has occurred to the probability of component combination B occurring independently; S203. Obtain the support (Support(A)) corresponding to component combination A, and select component combinations that are greater than or equal to the minimum support threshold Support_min as frequent items; for each pair of frequent items A and B, obtain the corresponding confidence and lift between them. If Confidence(A→B)≥Confidence_min and Lift(A→B)≥Lift_min, then mark the association between frequent items A and B as a strong association; obtain all frequent items marked as strongly associated, and search the historical formula data filtered in S202 based on the frequent items marked as strongly associated, and select the historical formula data with the most frequent items as the target historical formula data.
[0009] Furthermore, step S300 includes: S301. For each type of user, collect user evaluation information on the batch of beer corresponding to the target historical recipe data through an online platform. Based on the evaluation information of each user, construct a corresponding evaluation information vector H, where H=[h1,h2,...,hp], where h1 represents the evaluation information of the first dimension, h2 represents the evaluation information of the second dimension, and so on, with hp representing the evaluation information of the p-th dimension, where p represents the total number of dimensions of the evaluation information; calculate the average value avg_hq of the value corresponding to each dimension of the evaluation information vector H, where q takes values from 1 to p, and count the number of values M1, M2, and M3 of each dimension of the evaluation information that are greater than, equal to, or less than the average value. S302. Compare the magnitudes of the numerical values M1, M2, and M3, and select the evaluation information corresponding to the maximum value among M1, M2, and M3 as the overall evaluation information for the corresponding dimension. Summarize the overall evaluation information for each dimension of the evaluation information vector H to obtain the overall evaluation vector H'. If there are two or more equal numerical values among M1, M2, and M3 for a certain dimension, the evaluation information corresponding to M2 is selected as the overall evaluation information for that dimension. Obtain the target historical formula data for the corresponding category, extract the corresponding raw material data and corresponding raw material ratios from the target historical formula data, and identify the raw material data related to the overall evaluation information as the raw material data to be adjusted based on the overall evaluation vector H'. The identification of the raw material data to be adjusted is obtained by referring to the analysis method of "obtaining the component combination in the corresponding historical formula data based on user profile tags". Generate corresponding modification suggestions based on the raw material ratios of the raw material data to be adjusted based on the overall evaluation information.
[0010] Furthermore, step S400 includes: For each user category, suggestions for modifying the ingredient ratios in the corresponding target historical recipe data are output to relevant personnel. These personnel then modify the target historical recipe data accordingly, thereby obtaining personalized beer recipe data for the corresponding user category. Personalized beer is then produced based on this personalized beer recipe data. Feedback data from the corresponding user category on the personalized beer is collected using an online platform. This feedback data is analyzed in the same way as evaluation information, and relevant personnel make corresponding optimizations and adjustments based on the feedback data.
[0011] A personalized beer production management system based on artificial intelligence includes: a user demand collection and profile building module, a historical recipe data analysis module, a user evaluation analysis and recipe optimization module, and a personalized recipe production and feedback optimization module. The user demand collection and profile building module collects users' personalized beer demand information through an online platform, extracts corresponding demand features from the personalized beer demand information, classifies users according to demand features, and builds user profiles based on the classification results, thereby generating a personalized demand dataset. The historical recipe data analysis module collects beer purchase records of all users in the same category based on the personalized demand dataset, searches the recipe database based on the beer purchase records, obtains historical recipe data, and analyzes the correlation between different historical recipe data; based on the correlation analysis results, it obtains the target historical recipe data for the corresponding category of users; The user review analysis and formula optimization module collects evaluation information from users of the same category on the batches of beer corresponding to the target historical formula data through an online platform. Combining the evaluation information, it analyzes the ingredient ratios in the target historical formula data and obtains suggestions for modifying the ingredient ratios of the target historical formula data. The personalized recipe production and feedback optimization module outputs modification suggestions for the raw material ratios in the target historical recipe data to relevant personnel, who then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, personalized beer is produced, feedback data from the corresponding user category is collected, and the personalized beer recipe is optimized and adjusted based on the feedback data.
[0012] Furthermore, the personalized demand data collection and analysis module includes a demand information collection unit, a feature extraction and normalization unit, and a demand feature classification unit. The demand information collection unit collects users' personalized beer demand information through an online platform; the feature extraction and normalization unit performs feature extraction and normalization on the collected personalized beer demand information to construct the user's demand feature vector; the demand feature classification unit uses a hierarchical clustering algorithm to classify the user's demand feature vector, generate a user profile for each category, and output a personalized demand dataset.
[0013] Furthermore, the historical formula data analysis module includes a historical formula data collection and processing unit, a similarity calculation unit, and a correlation analysis unit; The historical recipe data collection and processing unit collects beer purchase records related to users of the same category based on the user demand dataset; it counts the purchase frequency and identifies the corresponding beer numbers based on the frequency, then searches for the historical recipe data corresponding to these beer numbers in the recipe database and extracts the raw material data of the historical recipes; the similarity calculation unit calculates the similarity between each historical recipe data based on the raw material component data in the historical recipes, forming a similarity matrix; the correlation analysis unit filters the historical recipe data based on the similarity matrix, analyzes the correlation between the component combinations in the filtered historical recipe data, and thus identifies the target historical recipe data for the corresponding category of users.
[0014] Furthermore, the user evaluation analysis and formula optimization module includes a user evaluation data collection and analysis unit and a formula adjustment suggestion generation unit; The user evaluation data collection and analysis unit collects user evaluation information on beer batches of the target historical recipe; constructs an evaluation vector based on the evaluation information; calculates the average value of each dimension, analyzes the numerical distribution of each dimension in the evaluation vector, and determines the overall evaluation information of each dimension; the recipe adjustment suggestion generation unit identifies raw material data related to the evaluation information based on the user evaluation analysis results and generates corresponding modification suggestions. The personalized formula production and feedback optimization module includes a personalized formula generation unit and a user feedback collection and analysis unit. The personalized recipe generation unit generates modification suggestions based on the recipe adjustment suggestion generation unit. Relevant personnel then adjust the corresponding target historical recipe data to obtain personalized beer recipe data. The user feedback collection and analysis unit produces personalized beer based on the personalized beer recipe data and collects feedback data from users of the corresponding categories. Based on the feedback data, relevant personnel make corresponding optimizations and adjustments.
[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: By collecting users' personalized needs information through an online platform and classifying users based on artificial intelligence algorithms (such as hierarchical clustering and association rule mining), it can accurately identify the needs and characteristics of each user, generate user profiles, and then provide customized beer recipes for different categories of users. This method can automatically select the recipes that best meet user needs from historical recipe databases and provide modification suggestions for the recipes, greatly reducing manual intervention and improving the efficiency and accuracy of recipe design. Through in-depth analysis of user feedback and evaluation information, the system can obtain modification suggestions for the ingredient ratio based on user evaluations of batches of beer. This process avoids the inefficiency of traditional manual adjustments, reduces unnecessary experimentation and adjustment costs, and ensures that the recipes better meet the personalized needs of users. Through in-depth analysis of a large amount of historical recipe data, this invention can uncover the potential correlations between recipes, identify frequently occurring ingredient combinations and their interrelationships. This correlation analysis can help manufacturers better understand the inherent patterns of historical recipes and provide a strong reference for future beer recipe development, reducing the trial-and-error costs of R&D. This invention continuously optimizes personalized beer recipes by collecting user feedback data on beer products and combining it with the results of evaluation information analysis. The closed-loop design of this feedback mechanism ensures that recipe adjustments are not only based on historical data but also respond in real time to changes in market and user needs, thereby achieving continuous product improvement and innovation. Attached Figure Description
[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1This is a schematic diagram of a personalized beer production management system module based on artificial intelligence according to the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Please see Figure 1 The present invention provides the following technical solution: A personalized beer production management system based on artificial intelligence includes: a user demand collection and profile building module, a historical recipe data analysis module, a user evaluation analysis and recipe optimization module, and a personalized recipe production and feedback optimization module. The user demand collection and profile building module collects users' personalized beer demand information through an online platform, extracts corresponding demand features from the personalized beer demand information, classifies users according to demand features, and builds user profiles based on the classification results, thereby generating a personalized demand dataset. The historical recipe data analysis module collects beer purchase records of all users in the same category based on the personalized demand dataset, searches the recipe database based on the beer purchase records, obtains historical recipe data, and analyzes the correlation between different historical recipe data; based on the correlation analysis results, it obtains the target historical recipe data for the corresponding category of users; The user review analysis and formula optimization module collects evaluation information from users of the same category on the batches of beer corresponding to the target historical formula data through an online platform. Combining the evaluation information, it analyzes the ingredient ratios in the target historical formula data and obtains suggestions for modifying the ingredient ratios of the target historical formula data. The personalized recipe production and feedback optimization module outputs modification suggestions for the raw material ratios in the target historical recipe data to relevant personnel, who then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, personalized beer is produced, feedback data from the corresponding user category is collected, and the personalized beer recipe is optimized and adjusted based on the feedback data.
[0019] The personalized demand data collection and analysis module includes a demand information collection unit, a feature extraction and normalization unit, and a demand feature classification unit. The demand information collection unit collects users' personalized beer demand information through an online platform; the feature extraction and normalization unit performs feature extraction and normalization on the collected personalized beer demand information to construct the user's demand feature vector; the demand feature classification unit uses a hierarchical clustering algorithm to classify the user's demand feature vector, generate a user profile for each category, and output a personalized demand dataset.
[0020] The historical formula data analysis module includes a historical formula data collection and processing unit, a similarity calculation unit, and a correlation analysis unit; The historical recipe data collection and processing unit collects beer purchase records related to users of the same category based on the user demand dataset; it counts the purchase frequency and identifies the corresponding beer numbers based on the frequency, then searches for the historical recipe data corresponding to these beer numbers in the recipe database and extracts the raw material data of the historical recipes; the similarity calculation unit calculates the similarity between each historical recipe data based on the raw material component data in the historical recipes, forming a similarity matrix; the correlation analysis unit filters the historical recipe data based on the similarity matrix, analyzes the correlation between the component combinations in the filtered historical recipe data, and thus identifies the target historical recipe data for the corresponding category of users.
[0021] The user evaluation analysis and formula optimization module includes a user evaluation data collection and analysis unit and a formula adjustment suggestion generation unit; The user evaluation data collection and analysis unit collects user evaluation information on beer batches of the target historical recipe; constructs an evaluation vector based on the evaluation information; calculates the average value of each dimension, analyzes the numerical distribution of each dimension in the evaluation vector, and determines the overall evaluation information of each dimension; the recipe adjustment suggestion generation unit identifies raw material data related to the evaluation information based on the user evaluation analysis results and generates corresponding modification suggestions. The personalized formula production and feedback optimization module includes a personalized formula generation unit and a user feedback collection and analysis unit. The personalized recipe generation unit generates modification suggestions based on the recipe adjustment suggestion generation unit. Relevant personnel then adjust the corresponding target historical recipe data to obtain personalized beer recipe data. The user feedback collection and analysis unit produces personalized beer based on the personalized beer recipe data and collects feedback data from users of the corresponding categories. Based on the feedback data, relevant personnel make corresponding optimizations and adjustments.
[0022] A personalized beer production management method based on artificial intelligence includes the following steps: Step S100. Collect users' personalized beer demand information through an online platform, extract corresponding demand features from the personalized beer demand information; classify users according to demand features, and construct user profiles based on the classification results, thereby generating a personalized demand dataset; Step S200. Based on the personalized demand dataset, collect the beer purchase records of all users in the same category, search the recipe database based on the beer purchase records, obtain historical recipe data, and analyze the correlation between different historical recipe data; based on the correlation analysis results, obtain the target historical recipe data for the corresponding category of users; Step S300. Collect evaluation information of the same category of users on the batches of beer corresponding to the target historical recipe data through the online platform, and analyze the raw material ratio in the target historical recipe data in combination with the evaluation information to obtain suggestions for modifying the raw material ratio of the target historical recipe data. Step S400: Output the suggested modifications to the raw material ratios in the target historical recipe data to relevant personnel, who will then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, produce personalized beer, collect feedback data from the corresponding user category on the personalized beer, and optimize and adjust the personalized beer recipe based on the feedback data.
[0023] Step S100 includes: S101. Collect users' personalized beer demand information through an online platform, wherein the online platform refers to a platform that collects users' personalized beer demand information and users' beer evaluation information; wherein the collection of users' personalized beer demand information is carried out through a questionnaire survey, thereby obtaining a quantitative representation of personalized beer demand information; extract the corresponding demand features from the personalized beer demand information and perform standardization processing to form a demand feature vector Xi, and Xi=[xi1,xi2,...,xin], where Xi represents the vector corresponding to the personalized beer demand information of the i-th user, i represents the user's data number; xi1 represents the first feature value in the personalized beer demand information of the i-th user, xi2 represents the second feature value in the personalized beer demand information of the i-th user, and so on, xin represents the nth feature value in the personalized beer demand information of the i-th user; n represents the number of feature values in the personalized beer demand information; In this embodiment, through a beer customization platform (e.g., a mobile application or website that specifically provides customized beer and collects user reviews of beer), the questionnaire may include, for example, the following: Flavor preferences: What is your preferred beer flavor? (e.g., 1. sweet, 2. sour, 3. neutral, etc.) Alcohol preference: What alcohol content of beer do you prefer? (e.g., 1. Below 5%, 2. 5%-7%, 3. Above 7%, etc.) Color preference: What is your preferred beer color? (e.g., 1. light, 2. dark, 3. amber, etc.) After each user completes the questionnaire, the platform will extract personalized beer demand characteristics based on their answers. For example, suppose the i-th user's questionnaire answers are as follows: Taste preference: Sweet (corresponding value: 1); Alcohol content preference: 5%-7% (corresponding value: 2); Color preference: Amber (corresponding value: 3); The platform transforms this information into a vector Xi=[xi1,xi2,xi3], where: xi1=1 indicates that the user prefers sweetness; xi2=2 indicates that the user prefers an alcohol content of 5%-7%; Xi3=3 indicates that the user prefers amber-colored beer.
[0024] S102. Based on the user's demand feature vector Xi, hierarchical clustering is used to classify the user's demand feature vector, thereby dividing the user into several categories; for each category, the mean value of the corresponding user on each feature dimension is calculated, thereby obtaining the demand feature mean vector Xa_c of the corresponding category, where c represents the category number; based on the demand feature mean vector Xa_c, user profile labels for the corresponding category are generated; based on each user's demand feature vector and user profile label, a personalized demand dataset G is constructed, and G={g1,g2,...,gm}, where g1 represents the demand feature vector and user profile label of the first user, g2 represents the demand feature vector and user profile label of the second user, and so on, gm represents the demand feature vector and user profile label of the m-th user, where m represents the user's data number.
[0025] Hierarchical clustering is used to classify user demand feature vectors, as detailed below: Calculate the distance d(Xi,Xj) between the demand feature vectors of each pair of users, and the specific calculation formula is as follows: d(Xi,Xj)={∑k∈[1,n],(xik-xjk) 2}^(1 / 2); Where Xi represents the demand feature vector of the i-th user, Xj represents the demand feature vector of the j-th user, xik represents the k-th demand feature value of the i-th user, and xjk represents the k-th demand feature value of the j-th user; k represents the dimension number of the demand feature vector, ranging from 1 to n; Initially, each user's demand feature vector Xi is set as an independent cluster, and there are m clusters in the initial stage. For each user, the two clusters with the smallest distance d(Xi,Xj) are selected and merged to form a new cluster, and the demand feature vector of the new cluster is the weighted average of all demand feature vectors of the two clusters. After merging the clusters, the distance between the new cluster and other clusters is recalculated. The aggregation process is repeated until the predetermined number of clusters is reached.
[0026] In this implementation, for two clusters A and B, assuming cluster A is the demand feature vector Xi and cluster B is the demand feature vector Xj, the demand feature vector XAB of the merged cluster is represented by the mean of the demand feature vectors of the users within the cluster: XAB=1 / (|A|+|B|)(∑i∈A,Xi+∑j∈B,Xj); Where |A| and |B| represent the number of users in cluster A and cluster B, respectively.
[0027] Step S200 includes: S201. Based on the personalized needs dataset, collect beer purchase records of all users in the same category; for each user in the corresponding category, calculate the purchase frequency of different beers based on the beer purchase records, arrange the purchase frequencies of different beers in descending order, and record the beer number with the highest purchase frequency as the pre-selection number; summarize the pre-selection numbers of all users in the corresponding category, search in the recipe database according to the pre-selection number, find the historical recipe data corresponding to the pre-selection number, and construct a historical recipe dataset; extract raw material data from the historical recipe data and construct a raw material dataset R, where R={r1,r2,...,rs], where r 1 represents the raw material data of the first type of historical formula data, r2 represents the raw material data of the second type of historical formula data, and so on, rs represents the raw material data of the s-th type of historical formula data; and each type of raw material data re represents a vector of its component characteristics, re=[ce1,ce2,...,cef], where ce1 represents the concentration of the first component in the raw material data of the e-th type of historical formula data, ce2 represents the concentration of the second component in the raw material data of the e-th type of historical formula data, and so on, cef represents the concentration of the f-th component in the raw material data of the e-th type of historical formula data, and f represents the number of components in the raw material; S202. Based on the raw material dataset, calculate the similarity of components among different historical formula data to obtain a similarity matrix S between different historical formulas. Here, Suv in the similarity matrix represents the similarity between raw material data ru and rv. Filter historical formula data in the similarity matrix that are greater than or equal to the similarity threshold S0. For the filtered historical formula data, use association rule mining to analyze the correlation between the corresponding historical formula data. The specific analysis process is as follows: Obtain user profile tags for the corresponding categories, and derive the corresponding ingredient combinations based on these tags. Calculate the support, confidence, and lift for different ingredient combinations in the historical recipe data. The formula for calculating support is: Support(A) = |A| / |D|, where |A| represents the frequency in the historical recipe dataset, and |D| represents the total number of recipes in the historical recipe dataset. The formula for calculating confidence is: Confidence(A→B)=Support(A∪B) / Support(A); Wherein, Confidence(A→B) represents the probability of component combination B occurring given that component combination A has occurred; Support(A∪B) represents the frequency with which component combinations A and B co-occur in the historical formula dataset; the formula for calculating the lift is: Lift(A→B)=Support(A∪B) / [Support(A)·Support(B)], where Lift(A→B) represents the ratio of the probability of component combination B occurring given that component combination A has occurred to the probability of component combination B occurring independently; In this embodiment, the corresponding component combinations are obtained based on user profile tags, as detailed below: Based on user preferences, each feature value (such as sweetness, alcohol content, and color) is mapped to a specific ingredient in the formula database. For example: Taste preferences (such as sweetness) may be directly related to certain ingredients (such as the proportion of sugar and malt). For example, a preference for "sweetness" may correspond to a higher sugar content or malt ratio.
[0028] Alcohol content preference (e.g., 5%-7%) is closely related to the fermentation process, yeast type, and sugar concentration in the raw materials. This preference may affect the yeast type and sugar content in the formula.
[0029] Color preferences (such as amber) are related to the degree and type of malt roasting. Amber usually means that darker roasted malt was used, so the color preference label will be mapped to the specific ingredient (such as dark malt).
[0030] By mapping and association rules, potential ingredient combinations are generated for each type of user profile label; for the example mentioned above, the user profile labels are [1,2,3], and the raw material data in the formula database is queried based on the characteristics of these labels.
[0031] For example: Taste preference: Sweet → Find historical recipes with higher sugar and malt content.
[0032] Alcohol content preference: 5%-7% → Find historical recipes with alcohol content in this range. You may need to find relevant data on fermentation temperature and yeast type.
[0033] Color preference: Amber → Look for recipes that use dark malt, which usually involves malt that has been roasted to a darker degree.
[0034] Once the user profile tags [1,2,3] are determined, the component combinations are generated in the following way: Retrieve corresponding historical recipes: Based on user profile tags, search the historical recipe database for recipes that match those characteristics. For example, select all recipes that are sweet, have an alcohol content between 5% and 7%, and are amber in color.
[0035] Extracting Ingredient Feature Vectors: Each historical recipe contains an ingredient feature vector, which represents the concentration or ratio of different ingredients (such as sugar, malt, yeast, etc.). For each historical recipe, the corresponding ingredient data (e.g., sugar concentration, yeast type, malt roasting level, etc.) is extracted.
[0036] Once multiple historical recipes and their component feature vectors are obtained, association rule mining (such as the Apriori algorithm) can be used to analyze the relationships between the recipe components: Support: Calculates the frequency of an ingredient combination in all formulations, reflecting how common the ingredient combination is.
[0037] Confidence: The probability that other combinations of components will occur given that a certain combination of components has occurred.
[0038] Lift: Measures the relative probability of other ingredient combinations occurring given a specific ingredient combination.
[0039] In this way, it is possible to determine which ingredient combinations best meet the user's personalized needs, thereby creating a more precise beer recipe for the user.
[0040] S203. Obtain the support (Support(A)) corresponding to component combination A, and select component combinations that are greater than or equal to the minimum support threshold Support_min as frequent items; for frequent items A and B, obtain the confidence and lift between them. If Confidence(A→B)≥Confidence_min and Lift(A→B)≥Lift_min, then mark frequent items A and B as strongly correlated frequent items; where Confidence_min and Lift_min represent the thresholds corresponding to confidence and lift, respectively; obtain all strongly correlated frequent items, and search the historical formula data filtered in S202 for strongly correlated frequent items, selecting the historical formula data with the most strongly correlated frequent items as the target historical formula data.
[0041] In this embodiment, assuming there are frequent items A, B, and C, the confidence and lift of frequent items A and B, frequent items A and C, and frequent items B and C are calculated respectively. Assuming they are compared with the thresholds corresponding to the confidence and lift, the strongly correlated frequent items A, B, and C are obtained. Based on the strongly correlated frequent items A, B, and C, the selected historical formula data are searched. Assuming there is a historical formula data with the most strongly correlated frequent items, namely A and B, this historical formula data is used as the target historical formula data.
[0042] Step S300 includes: S301. For each type of user, collect user evaluation information on the batch of beer corresponding to the target historical recipe data through an online platform. Based on the evaluation information of each user, construct a corresponding evaluation information vector H, where H=[h1,h2,...,hp], where h1 represents the evaluation information of the first dimension, h2 represents the evaluation information of the second dimension, and so on, with hp representing the evaluation information of the p-th dimension, where p represents the total number of dimensions of the evaluation information; calculate the average value avg_hq of the value corresponding to each dimension of the evaluation information vector H, where q takes values from 1 to p, and count the number of values M1, M2, and M3 of each dimension of the evaluation information that are greater than, equal to, or less than the average value. S302. Compare the magnitudes of the numerical values M1, M2, and M3, and select the evaluation information corresponding to the maximum value among M1, M2, and M3 as the overall evaluation information for the corresponding dimension. Summarize the overall evaluation information for each dimension of the evaluation information vector H to obtain the overall evaluation vector H'. If there are two or more equal numerical values among M1, M2, and M3 for a certain dimension, the evaluation information corresponding to M2 is selected as the overall evaluation information for that dimension. Obtain the target historical formula data for the corresponding category, extract the corresponding raw material data and corresponding raw material ratios from the target historical formula data, and identify the raw material data related to the overall evaluation information as the raw material data to be adjusted based on the overall evaluation vector H'. The identification of the raw material data to be adjusted is obtained by referring to the analysis method of "obtaining the component combination in the corresponding historical formula data based on user profile tags". Generate corresponding modification suggestions based on the raw material ratios of the raw material data to be adjusted based on the overall evaluation information.
[0043] In this embodiment, users rate batches of beer corresponding to the target historical recipe through an online platform. Each user's rating includes the following dimensions (for example): Taste evaluation: including dimensions such as sweetness, bitterness, and sourness; Alcohol content rating: Users give a rating based on their perception of alcohol concentration; Color evaluation: Scoring of the beer's appearance characteristics such as color and clarity; Each user will assign a numerical rating (e.g., from 1 to 5) for these dimensions. Taking sweetness as an example, where the sweetness rating is strong, 1 indicates very low or almost no sweetness; 2 indicates low sweetness but still has some sweetness; 3 indicates moderate sweetness; 4 indicates high sweetness; and 5 indicates very high sweetness. Other dimensions follow the same pattern, forming an evaluation information vector H. For a specific user, assuming the evaluation dimension is p (e.g., p=3), the user's evaluation information vector H is: [h1, h2, h3], where: h1: Taste rating, h2: Alcohol content rating, h3: Color rating; For example: User A's rating vector: H=[5,5,3] User B's evaluation vector: H=[5,4,5] User C's evaluation vector: H=[2,4,4]; For each dimension, the following three types of quantities are counted: M1: The number of users whose rating is higher than the average of this dimension. M2: The number of users whose rating equals the average of this dimension. M3: The number of users whose rating is lower than the average of this dimension; Taking h1 as an example of taste score, assuming it is sweet, we know from the above data that the average sweetness score corresponding to h1 is avg_h1=4; therefore, M1=2, M2=0, M3=1. Since M1>M3>M2, that is, the taste evaluation is "very sweet", then the proportion of ingredients related to sweetness needs to be appropriately reduced to reduce the sweetness. Therefore, the modification suggestion is to reduce the proportion of ingredients related to sweetness.
[0044] Step S400 includes: For each user category, suggestions for modifying the ingredient ratios in the corresponding target historical recipe data are output to relevant personnel. These personnel then modify the target historical recipe data accordingly, thereby obtaining personalized beer recipe data for the corresponding user category. Personalized beer is then produced based on this personalized beer recipe data. Feedback data from the corresponding user category on the personalized beer is collected using an online platform. This feedback data is analyzed in the same way as evaluation information, and relevant personnel make corresponding optimizations and adjustments based on the feedback data.
[0045] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0046] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A personalized customization management method for beer production based on artificial intelligence, characterized in that: The method includes the following steps: Step S100. Collect users' personalized beer demand information through an online platform, extract corresponding demand features from the personalized beer demand information; classify users according to demand features, and construct user profiles based on the classification results, thereby generating a personalized demand dataset; Step S200. Based on the personalized demand dataset, collect the beer purchase records of all users in the same category, search the recipe database based on the beer purchase records, obtain historical recipe data, and analyze the correlation between different historical recipe data; based on the correlation analysis results, obtain the target historical recipe data for the corresponding category of users; Step S300. Collect evaluation information of the same category of users on the batches of beer corresponding to the target historical recipe data through the online platform, and analyze the raw material ratio in the target historical recipe data in combination with the evaluation information to obtain suggestions for modifying the raw material ratio of the target historical recipe data. Step S400: Output the suggested modifications to the raw material ratios in the target historical recipe data to relevant personnel, who will then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, produce personalized beer, collect feedback data from the corresponding user category on the personalized beer, and optimize and adjust the personalized beer recipe based on the feedback data.
2. The method for personalized customization management of beer production based on artificial intelligence according to claim 1, characterized in that: Step S100 includes: S101. Collect users' personalized beer demand information through an online platform, wherein the online platform refers to a platform that collects users' personalized beer demand information and users' beer evaluation information; wherein the collection of users' personalized beer demand information is carried out through a questionnaire survey, thereby obtaining a quantitative representation of personalized beer demand information; extract the corresponding demand features from the personalized beer demand information and perform standardization processing to form a demand feature vector Xi, and Xi=[xi1,xi2,...,xin], where Xi represents the vector corresponding to the personalized beer demand information of the i-th user, i represents the user's data number; xi1 represents the first feature value in the personalized beer demand information of the i-th user, xi2 represents the second feature value in the personalized beer demand information of the i-th user, and so on, xin represents the nth feature value in the personalized beer demand information of the i-th user; n represents the number of feature values in the personalized beer demand information; S102. Based on the user's demand feature vector Xi, hierarchical clustering is used to classify the user's demand feature vector, thereby dividing the user into several categories; for each category, the mean value of the corresponding user on each feature dimension is calculated, thereby obtaining the demand feature mean vector Xa_c of the corresponding category, where c represents the category number; based on the demand feature mean vector Xa_c, user profile labels for the corresponding category are generated; based on each user's demand feature vector and user profile label, a personalized demand dataset G is constructed, and G={g1,g2,...,gm}, where g1 represents the demand feature vector and user profile label of the first user, g2 represents the demand feature vector and user profile label of the second user, and so on, gm represents the demand feature vector and user profile label of the m-th user, where m represents the user's data number.
3. The method for personalized customization management of beer production based on artificial intelligence according to claim 2, characterized in that: The use of hierarchical clustering to classify user demand feature vectors is detailed below: Calculate the distance d(Xi,Xj) between the demand feature vectors of each pair of users, and the specific calculation formula is as follows: d(Xi,Xj)={∑k∈[1,n],(wing-xjk) 2 }^(1 / 2): Where Xi represents the demand feature vector of the i-th user, Xj represents the demand feature vector of the j-th user, xik represents the k-th demand feature value of the i-th user, and xjk represents the k-th demand feature value of the j-th user; k represents the dimension number of the demand feature vector, ranging from 1 to n; Initially, each user's demand feature vector Xi is set as an independent cluster, and there are m clusters in the initial stage. For each user, the two clusters with the smallest distance d(Xi,Xj) are selected and merged to form a new cluster, and the demand feature vector of the new cluster is the weighted average of all demand feature vectors of the two clusters. After merging the clusters, the distance between the new cluster and other clusters is recalculated. The aggregation process is repeated until the predetermined number of clusters is reached.
4. The method for personalized customization management of beer production based on artificial intelligence according to claim 2, characterized in that: Step S200 includes: S201. Based on the personalized needs dataset, collect beer purchase records of all users in the same category; for each user in the corresponding category, calculate the purchase frequency of different beers based on the beer purchase records, arrange the purchase frequencies of different beers in descending order, and record the beer number with the highest purchase frequency as the pre-selection number; summarize the pre-selection numbers of all users in the corresponding category, search in the recipe database according to the pre-selection number, find the historical recipe data corresponding to the pre-selection number, and construct a historical recipe dataset; extract raw material data from the historical recipe data and construct a raw material dataset R, where R={r1,r2,...,rs], where r 1 represents the raw material data of the first type of historical formula data, r2 represents the raw material data of the second type of historical formula data, and so on, rs represents the raw material data of the s-th type of historical formula data; and each type of raw material data re represents a vector of its component characteristics, re=[ce1,ce2,...,cef], where ce1 represents the concentration of the first component in the raw material data of the e-th type of historical formula data, ce2 represents the concentration of the second component in the raw material data of the e-th type of historical formula data, and so on, cef represents the concentration of the f-th component in the raw material data of the e-th type of historical formula data, and f represents the number of components in the raw material; S202. Based on the raw material dataset, calculate the similarity of components among different historical formula data to obtain a similarity matrix S between different historical formulas. Here, Suv in the similarity matrix represents the similarity between raw material data ru and rv. Filter historical formula data in the similarity matrix that are greater than or equal to the similarity threshold S0. For the filtered historical formula data, use association rule mining to analyze the correlation between the corresponding historical formula data. The specific analysis process is as follows: Obtain user profile tags for the corresponding categories, and derive the corresponding ingredient combinations based on these tags. Calculate the support, confidence, and lift for different ingredient combinations in the historical recipe data. The formula for calculating support is: Support(A) = |A| / |D|, where |A| represents the frequency in the historical recipe dataset, and |D| represents the total number of recipes in the historical recipe dataset. The formula for calculating confidence is: Confidence(A→B)=Support(A∪B) / Support(A); Wherein, Confidence(A→B) represents the probability of component combination B occurring given that component combination A has occurred; Support(A∪B) represents the frequency with which component combinations A and B co-occur in the historical formula dataset; the formula for calculating the lift is: Lift(A→B)=Support(A∪B) / [Support(A)·Support(B)], where Lift(A→B) represents the ratio of the probability of component combination B occurring given that component combination A has occurred to the probability of component combination B occurring independently; S203. Obtain the support (Support(A)) corresponding to component combination A, and select component combinations that are greater than or equal to the minimum support threshold Support_min as frequent items; for each pair of frequent items A and B, obtain the corresponding confidence and lift between them. If Confidence(A→B)≥Confidence_min and Lift(A→B)≥Lift_min, then mark the association between frequent items A and B as a strong association; obtain all frequent items marked as strongly associated, and search the historical formula data filtered in S202 based on the frequent items marked as strongly associated, and select the historical formula data with the most frequent items as the target historical formula data.
5. The method for personalized customization management of beer production based on artificial intelligence according to claim 4, characterized in that: Step S300 includes: S301. For each type of user, collect user evaluation information on the batch of beer corresponding to the target historical recipe data through an online platform. Based on the evaluation information of each user, construct a corresponding evaluation information vector H, where H=[h1,h2,...,hp], where h1 represents the evaluation information of the first dimension, h2 represents the evaluation information of the second dimension, and so on, with hp representing the evaluation information of the p-th dimension, where p represents the total number of dimensions of the evaluation information; calculate the average value avg_hq of the value corresponding to each dimension of the evaluation information vector H, where q takes values from 1 to p, and count the number of values M1, M2, and M3 of each dimension of the evaluation information that are greater than, equal to, or less than the average value. S302. Compare the magnitudes of the numerical values M1, M2, and M3, and select the evaluation information corresponding to the maximum value among the numerical values M1, M2, and M3 as the overall evaluation information for the corresponding dimension. Summarize the overall evaluation information for each dimension of the evaluation information vector H to obtain the overall evaluation vector H'. If there are two or more equal numerical values among M1, M2, and M3 for a certain dimension, the evaluation information corresponding to M2 is selected as the overall evaluation information for that dimension. Obtain the target historical formula data for the corresponding category, extract the corresponding raw material data and corresponding raw material ratios from the target historical formula data, identify the raw material data related to the overall evaluation information as the raw material data to be adjusted based on the overall evaluation vector H', and generate corresponding modification suggestions based on the raw material ratios of the raw material data to be adjusted.
6. The method for personalized customization management of beer production based on artificial intelligence according to claim 5, characterized in that: Step S400 includes: For each user category, suggestions for modifying the ingredient ratios in the corresponding target historical recipe data are output to relevant personnel. These personnel then modify the target historical recipe data accordingly, thereby obtaining personalized beer recipe data for the corresponding user category. Personalized beer is then produced based on this personalized beer recipe data. Feedback data from the corresponding user category on the personalized beer is collected using an online platform. This feedback data is analyzed in the same way as evaluation information, and relevant personnel make corresponding optimizations and adjustments based on the feedback data.
7. An AI-based personalized beer production management system, applied to any one of the AI-based personalized beer production management methods according to claims 1-6, characterized in that: The system includes: a user needs collection and profile building module, a historical formula data analysis module, a user evaluation analysis and formula optimization module, and a personalized formula production and feedback optimization module; The user demand collection and profile building module collects users' personalized beer demand information through an online platform, extracts corresponding demand features from the personalized beer demand information, classifies users according to demand features, and builds user profiles based on the classification results, thereby generating a personalized demand dataset. The historical recipe data analysis module collects beer purchase records of all users in the same category based on the personalized demand dataset, searches the recipe database based on the beer purchase records, obtains historical recipe data, and analyzes the correlation between different historical recipe data; based on the correlation analysis results, it obtains the target historical recipe data for the corresponding category of users. The user evaluation analysis and formula optimization module collects evaluation information from users of the same category on the batches of beer corresponding to the target historical formula data through an online platform. Combining the evaluation information, it analyzes the raw material ratio in the target historical formula data and obtains suggestions for modifying the raw material ratio of the target historical formula data. The personalized recipe production and feedback optimization module outputs modification suggestions for the raw material ratios in the target historical recipe data to relevant personnel, who then make the corresponding modifications to obtain personalized beer recipe data for the corresponding user category. Based on the generated personalized beer recipe, personalized beer is produced, feedback data from the corresponding user category is collected, and the personalized beer recipe is optimized and adjusted based on the feedback data.
8. The personalized beer production management system based on artificial intelligence according to claim 7, characterized in that: The personalized demand data collection and analysis module includes a demand information collection unit, a feature extraction and normalization unit, and a demand feature classification unit. The demand information collection unit collects users' personalized beer demand information through an online platform; the feature extraction and normalization unit performs feature extraction and normalization processing on the collected personalized beer demand information to construct a user demand feature vector; The demand feature classification unit uses a hierarchical clustering algorithm to classify the user's demand feature vectors, generate a user profile for each category, and output a personalized demand dataset.
9. The personalized beer production management system based on artificial intelligence according to claim 7, characterized in that: The historical formula data analysis module includes a historical formula data collection and processing unit, a similarity calculation unit, and a correlation analysis unit; The historical recipe data collection and processing unit collects beer purchase records related to users of the same category based on the user demand dataset; The purchase frequency is counted and the corresponding beer number is identified based on the frequency. Then, the historical recipe data corresponding to these beer numbers is searched from the recipe database, and the raw material data of the historical recipe is extracted. The similarity calculation unit calculates the similarity between various historical formula data based on the raw material composition data in the historical formulas, forming a similarity matrix; The correlation analysis unit filters historical formula data based on a similarity matrix, analyzes the correlation between component combinations in the filtered historical formula data, and thus identifies the target historical formula data of the corresponding user category.
10. The personalized beer production management system based on artificial intelligence according to claim 7, characterized in that: The user evaluation analysis and formula optimization module includes a user evaluation data collection and analysis unit and a formula adjustment suggestion generation unit; The user evaluation data collection and analysis unit collects user evaluation information on beer batches of the target historical recipe; constructs an evaluation vector based on the evaluation information; calculates the average value of each dimension, analyzes the numerical distribution of each dimension in the evaluation vector, and determines the overall evaluation information of each dimension; the recipe adjustment suggestion generation unit identifies raw material data related to the evaluation information based on the user evaluation analysis results and generates corresponding modification suggestions. The personalized formula production and feedback optimization module includes a personalized formula generation unit and a user feedback collection and analysis unit. The personalized recipe generation unit adjusts the corresponding target historical recipe data based on the modification suggestions generated by the recipe adjustment suggestion generation unit, thereby obtaining personalized beer recipe data. The user feedback collection and analysis unit produces personalized beer based on personalized beer recipe data, collects feedback data from users of the corresponding categories on personalized beer, and makes corresponding optimizations and adjustments based on the feedback data.