Baijiu after-drinking comfort level prediction system and method based on multi-stage tasting data
The white wine after-drinking comfort prediction system, constructed through multi-stage data collection and machine learning algorithms, solves the problem of difficulty in quantifying after-drinking comfort in existing technologies, realizes personalized early warning and process optimization, and improves the scientificity and accuracy of the white wine evaluation system.
Patent Information
- Application Number
- CN202510804267.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-09-16
AI Technical Summary
The existing liquor evaluation system fails to effectively predict the comfort level after drinking, lacks consideration of individual differences, and is unable to integrate multi-stage data for dynamic correlation analysis, resulting in low accuracy in drunkenness risk warning and post-drinking comfort level prediction, and difficulty in accurately locating the root cause of the problem.
Through a multi-stage tasting data acquisition module, combined with high-precision sensors and wearable devices, we collect information about the wine before drinking, behavioral data during drinking, and physiological indicators after drinking. We use machine learning algorithms to build a prediction model to achieve a quantitative assessment of the comfort level after drinking liquor.
It has achieved a scientific and quantitative assessment of the comfort level after drinking liquor, provided personalized drinking risk warnings and process optimization guidance, and improved consumer satisfaction and product competitiveness.
Smart Images
Figure SMS_1 
Figure SMS_2 
Figure SMS_3
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of liquor tasting, and specifically proposes a liquor drinking comfort prediction system and method based on multi-stage tasting data. Background Art
[0002] As a traditional Chinese beverage, the quality assessment of baijiu has long relied on pre-consumption physical and chemical analysis (such as alcohol content, total acidity, total esters, esters, aldehydes, and higher alcohols) and sensory evaluation by professional tasters before or during consumption (such as color, aroma, mouthfeel, and style). While this evaluation method can reflect the basic characteristics of baijiu to a certain extent, it has significant limitations: Existing technologies primarily focus on static evaluations before or during the initial stages of consumption, lacking effective predictive tools for post-drink comfort, a common concern for consumers (e.g., headache, dry mouth, nausea, gastrointestinal discomfort, and next-day sensations). Post-drink discomfort is a key factor influencing consumer satisfaction and repurchase intention, yet it is overlooked by traditional evaluation systems due to its difficulty in quantifying.
[0003] Existing methods fail to effectively integrate and dynamically correlate data from different stages. Specifically, Pre-drinking data (wine body information) is disconnected from in-drinking data (actual drinking behavior, such as drinking speed, amount consumed, the impact of food pairing, and dynamic changes in taste). There is also a lack of systematic causal or correlation analysis between pre-drinking and post-drinking data (physiological responses and subjective discomfort feedback). This data silo phenomenon makes it impossible to capture key dynamic changes throughout the drinking process and their impact on ultimate comfort.
[0004] Existing assessment systems are typically based on group averages or standard conditions, failing to fully account for individual differences. Consumers' physical characteristics (such as weight, age, gender, and underlying health status), drinking habits (drinking frequency, drinking speed, and alcohol tolerance), and immediate state (such as fatigue) significantly influence post-drinking reactions. The lack of integration of these individual factors significantly reduces the accuracy of intoxication risk warnings and post-drinking comfort predictions, making it difficult to provide truly personalized drinking guidance.
[0005] When consumers report experiencing discomfort after drinking, existing technologies struggle to pinpoint the root cause. Due to a lack of systematic correlation analysis between post-drink physiological data and wine composition (particularly trace substances that may trigger discomfort, such as fusel oils and aldehydes), manufacturers face an optimization blind spot when it comes to product improvements. The inability to establish a quantitative relationship between wine composition and post-drink comfort results in a lack of accurate data support for process improvements and recipe optimization, leading to low efficiency.
[0006] Therefore, the liquor industry is in urgent need of a technical solution that can overcome the above-mentioned defects. Summary of the Invention
[0007] In view of this, the present invention proposes a system and method for predicting the comfort level after drinking liquor based on multi-stage tasting data.
[0008] The technical solution of the present invention is achieved as follows: The present invention provides a system for predicting the comfort level of liquor after drinking based on multi-stage tasting data, comprising: The data collection module is used to collect data from the entire liquor tasting cycle in stages, including: Before drinking: enter the basic information of the liquor, including the chromatographic composition of the liquor; During drinking: High-precision sensors monitor drinking speed and amount in real time, and simultaneously record the taster's numerical score of taste changes; After drinking: wearable devices are used to collect the user's physiological data such as heart rate, blood pressure, and body temperature, and collect the user's subjective discomfort feedback; A data processing center, connected to the data processing center, is used to: Abnormal data of cleaning equipment; Store data by drinking stage and build a structured database; Use clustering or regression algorithms to conduct association analysis, and use cluster analysis (such as K-means) to explore the key factor combinations that affect post-drink comfort; The prediction model building module is connected to the data processing center and is used to: Based on the machine learning algorithm, a prediction model is constructed with multi-stage characteristic variables as input and post-drinking comfort score as output; Optimize model generalization capabilities through cross-validation and parameter tuning; Output quantitative comfort score and probability description of discomfort symptoms; The user interaction interface is used to provide consumers, tasters and manufacturers with differentiated function access portals, and supports real-time feedback of prediction results on multiple terminals.
[0009] In some embodiments, the high-precision sensor is a wearable flow monitoring device that tracks drinking dynamic data in real time and synchronizes it to a data processing center; the wearable device is a heart rate belt or smart bracelet with continuous monitoring function, which continuously collects physiological indicator data for ≥60 minutes, and the user feedback questionnaire is pushed in real time through the interactive interface and is forcibly triggered to be submitted within 60 minutes after the physiological data collection is completed.
[0010] In some embodiments, the feature association analysis performed by the data processing center includes: Identify the correlation between "high alcohol content + fast drinking speed" and "increased heart rate + headache"; The causal relationship between fusel oil content in wine and the degree of dry mouth after drinking was quantified by Granger causality test.
[0011] In some embodiments, the machine learning algorithm used in the prediction model building module is a support vector machine or a neural network; the model output includes a 0-10 point comfort quantification score and the probability of occurrence of specific discomfort symptoms, and the parameter tuning is achieved using grid search or Bayesian optimization.
[0012] In some embodiments, the user interface includes: Consumer side: Enter liquor information to obtain a personalized comfort level prediction report; Wine taster side: input professional sensory evaluation and mark data; Manufacturer side: Check the post-drinking comfort analysis results of batch products and optimize the production process.
[0013] In some embodiments, the data processing center further comprises: Data standardization unit, which normalizes multi-source heterogeneous data; Feature screening unit, which screens significant influencing factors using random forest or LASSO regression algorithms.
[0014] In a second aspect, the present invention further provides a method for predicting the comfort level of liquor after drinking based on multi-stage tasting data. The method is used in the above-mentioned prediction system and comprises the following steps: (1) System initialization: perform self-tests on sensors and wearable devices, configure data sampling frequency, and clear historical cache; (2) Multi-stage data collection: Scan the code or manually enter the basic information of the liquor before drinking; During drinking, sensors monitor drinking behavior data in real time and record taster scores; After drinking, physiological data will be collected for ≥60 minutes and a user feedback questionnaire will be sent; (3) Data processing and analysis: Normalize key feature variables; The significant influencing factors are extracted through the feature screening algorithm to generate a feature matrix containing ≥10-dimensional features; (4) Prediction model operation: The feature matrix is input into the trained prediction model, and the quantitative score of post-drinking comfort and description of discomfort symptoms are output; (5) Result feedback and application: Real-time feedback of results to users to guide consumers’ drinking decisions, optimize tasters’ evaluation standards, and improve manufacturers’ processes.
[0015] In some embodiments, the feature screening in step (3) includes: Random forest is used to calculate feature importance ranking; Redundant variables were eliminated through LASSO regression, and the factor combinations significantly correlated with post-drinking comfort were retained.
[0016] In some embodiments, the output of the prediction model in step (4) is in the form of: Comfort score: 0-10 point scale, with higher scores indicating better comfort; Description of discomfort symptoms: Output the probability of occurrence of headache, dry mouth, and gastrointestinal discomfort in the form of probability.
[0017] In some implementations, the manufacturer-side application in step (5) specifically includes: Based on the comfort rating of batch products after drinking, locate the threshold of wine components that cause discomfort; According to the quantitative relationship between the key factor combination and the discomfort symptoms, the raw materials and koji, brewing process and aging conditions are adjusted based on the quantitative relationship of the key factor combination.
[0018] The manufacturer generates process adjustment parameter recommendations based on the comfort scores of batch products and the corresponding wine composition data.
[0019] The present invention has the following beneficial effects compared to the prior art: The present invention integrates multi-stage tasting data such as pre-drinking wine body information, drinking behavior data and dynamic taste scoring, post-drinking physiological indicators and subjective feedback, and combines it with machine learning algorithms to build a prediction model, thereby realizing for the first time a scientific quantitative evaluation of the comfort level after drinking liquor. This technology breaks through the limitations of traditional static evaluation, and can not only provide consumers with personalized drinking risk warnings and comfort prediction reports, but also accurately guide manufacturers to optimize process parameters based on the quantitative correlation model between wine body components and discomfort symptoms, thereby filling the industry's technical gap in the field of post-drinking comfort prediction, and promoting the upgrade of the liquor quality evaluation system from "physical and chemical indicators before drinking" to "full-cycle health experience", and comprehensively improving product competitiveness and consumer satisfaction. DETAILED DESCRIPTION
[0020] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described 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 making creative efforts are within the scope of protection of the present invention.
[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which embodiments of the present invention belong. If the definitions set forth in this section are contrary to or otherwise inconsistent with definitions set forth in the patents, patent applications, published patent applications, and other publications incorporated herein by reference, the definitions listed in this section take precedence over the definitions incorporated herein by reference.
[0022] Unless otherwise specified, the methods used in the following examples are conventional methods. The materials, reagents, and instruments used are conventional materials, reagents, and instruments in the art, unless otherwise specified, and can be obtained commercially by those skilled in the art.
[0023] When an amount, concentration or other value or parameter is expressed as a range, a preferred range or a range defined by a series of upper preferred values and lower preferred values, this should be understood as specifically disclosing all ranges formed by any pairing of any upper range limit or preferred value with any lower range limit or preferred value, regardless of whether the range is disclosed alone. For example, when a range "1 to 5" is disclosed, the described range should be interpreted as including the ranges "1 to 4", "1 to 3", "1 to 2", "1 to 2 and 4 to 5", "1 to 3 and 5", etc. When a numerical range is described herein, unless otherwise stated, the range is intended to include its endpoints and all integers and fractions within the range. In the present specification and claims, range definitions may be combined and / or interchanged, and if not otherwise stated, such ranges include all subranges contained therein.
[0024] System hardware configuration Data acquisition equipment: Data entry before drinking: The mobile terminal calls the National Liquor Quality Inspection Database API (interface standard GB / T10781.2-2023) to obtain the chromatographic composition information of the liquor; Monitoring during drinking: Smart coaster (model X-CupPad, pressure sensor range 0-500g, accuracy ±0.1g), real-time calculation of drinking speed (mL / s) and single drinking volume (mL); The tablet terminal records the taster's taste scores. The scoring criteria are shown in the following table:
[0025] Monitoring after drinking: Huawei GT4 smart wristband (heart rate monitoring error ≤±3bpm, sampling frequency 1Hz), continuously collecting physiological data for ≥60 minutes; The user feedback questionnaire is pushed through the APP, and the screen is forced to lock until submission (a timeout of 30 minutes will trigger a text message reminder).
[0026] Prediction model training method Training set construction: Data source: 300 participants (male:female ratio 1:1, age 25-60 years, BMI 18.5-28.0, all signed informed consent) tasted 12 types of baijiu (alcohol content 38%-65%). Feature Engineering Process: 1. Generate composite features: Alcohol intake per unit body weight (g / kg) = total alcohol intake (g) / body weight (kg) Spicy impact index = Spicy peak score × drinking speed (mL / s) 2. Feature screening (LASSO regression, regularization strength α=0.01): Fifteen core features (importance > 0.01) were retained, including fusel oil content, average heart rate 30 minutes after drinking, and spicy impact index; Model building Network structure: 3-layer fully connected neural network (15 nodes in the input layer, 10 nodes in the hidden layer, ReLU activation; 3 nodes in the output layer corresponding to comfort level probabilities); Parameter optimization: Grid search determines the optimal parameter combination, see the table below:
[0027] Performance verification: 10-fold cross validation accuracy is 91.2% and recall is 89.4%.
[0028] Example: Prediction of drinking comfort and process optimization of Maotai-flavor liquor Test Subject: Liquor: 53% vol sauce-flavor liquor (batch 2023-07, fusel oil content: n-propanol 1018 mg / L, isobutanol 320 mg / L); User: Male, 35 years old, weight 70 kg, drinking habit: moderate (2 times a week).
[0029] Implementation steps Data collection: Before drinking: Scan the code to obtain wine information; During drinking: 1. Coaster monitoring: drinking speed 0.8mL / s, single drinking volume 15mL, total intake 120mL (alcohol volume 50g); 2. Sommelier rating: Spicy peak score 7.2 (lasting 8 seconds), aroma layering score 8.5; After drinking: 1. Bracelet monitoring: heart rate increased by 15%, body temperature increased by 0.3℃; 2. Questionnaire feedback: dry mouth degree 3 / 5 points; Model predictions: Input features: Alcohol content 53, Spicy impact index 5.76, n-propanol 1018mg / L, isobutanol 320mg / L, Δ heart rate +15%, Aroma layering 8.5, Alcohol content per unit body weight 0.71g / kg, Δ body temperature +0.3°C, Standard deviation of drinking interval 12s, Food pairing type 0, Tolerance 1 1. `Alcohol content` (from wine body information) 2. Spicy Impact Index (= Peak Spicy Score × Drinking Speed, derived from taster scores and sensor) 3. Fusel oil content (n-propanol, from wine composition testing) 4. `Δ Heart rate compared to baseline value` (average 30 minutes after drinking, from wearable device) 5. Aroma layering score (from taster ratings) 6. Alcohol per unit body weight (g / kg) (= total alcohol intake / body weight) 7. Change in body temperature after drinking (℃) 8. `Standard deviation of drinking interval (seconds)` (reflects drinking rhythm) 9. `Meal type code` (0-light, 1-greasy, from user input) 10. `User's alcohol tolerance` (0 - low, 1 - medium, 2 - high, from user profile) Output: Comfort rating: 6.8 / 10 Symptom probability: dry mouth 68%, headache 12% Recommendation: Single drinking volume ≤ 100mL, drinking speed < 0.5mL / s.
[0030] Manufacturer process optimization: Problem location: Analyzing 100 samples from the same batch, the average comfort score was 5.9 points → n-propanol > 100 mg / L causes a > 75% probability of dry mouth. Parameter adjustment: Extend high temperature pile fermentation time: 48h → 72h; Increase the rate of head interception: 30% → 35%; The verification results are shown in the following table:
[0031] Individual metabolic differences result in sensitive people still experiencing dry mouth reactions at 85 mg / L.
[0032] Implementation effect verification: The results of the comparison of 100 users’ measured data with the traditional method are shown in the table below:
[0033] 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 scope of protection of the present invention.
Claims
1. A liquor drinking comfort prediction system based on multi-stage tasting data, characterized by: include: The data collection module is used to collect data from the entire liquor tasting cycle in stages, including: Before drinking: enter the basic information of the liquor, including the chromatographic composition of the liquor; During drinking: High-precision sensors monitor drinking speed and amount in real time, and simultaneously record the taster's numerical score of taste changes; After drinking: wearable devices are used to collect the user's physiological data such as heart rate, blood pressure, and body temperature, and collect the user's subjective discomfort feedback; A data processing center, connected to the data processing center, is used to: Abnormal data of cleaning equipment; Store data by drinking stage and build a structured database; Use clustering or regression algorithms to conduct association analysis and explore the key factor combinations that affect post-drink comfort; The prediction model building module is connected to the data processing center and is used to: Based on the machine learning algorithm, a prediction model is constructed with multi-stage characteristic variables as input and post-drinking comfort score as output; Optimize model generalization capabilities through cross-validation and parameter tuning; Output quantitative comfort score and probability description of discomfort symptoms; The user interaction interface is used to provide consumers, tasters and manufacturers with differentiated function access portals, and supports real-time feedback of prediction results on multiple terminals.
2. The liquor drinking comfort prediction system based on multi-stage tasting data according to claim 1, characterized in that: The high-precision sensor is a wearable flow monitoring device that tracks drinking dynamic data in real time and synchronizes it to the data processing center; the wearable device is a heart rate belt or smart bracelet with continuous monitoring function, which continuously collects physiological indicator data for ≥60 minutes.
3. The liquor drinking comfort prediction system based on multi-stage tasting data according to claim 1, characterized in that: The feature association analysis performed by the data processing center includes: Identify the correlation between "high alcohol content + fast drinking speed" and "increased heart rate + headache"; To quantify the causal relationship between fusel oil content in wine and the degree of dry mouth after drinking.
4. The liquor drinking comfort prediction system based on multi-stage tasting data according to claim 1, characterized in that: The machine learning algorithm used in the prediction model construction module is a support vector machine or a neural network; the model output includes a 0-10 point comfort quantification score and the probability of occurrence of specific discomfort symptoms.
5. The liquor drinking comfort prediction system based on multi-stage tasting data according to claim 1 is characterized in that: The user interaction interface includes: Consumer side: Enter liquor information to obtain a personalized comfort level prediction report; Wine taster side: input professional sensory evaluation and mark data; Manufacturer side: Check the post-drinking comfort analysis results of batch products and optimize the production process.
6. The liquor drinking comfort prediction system based on multi-stage tasting data according to claim 1, characterized in that: The data processing center further comprises: Data standardization unit, which normalizes multi-source heterogeneous data; Feature screening unit, which screens significant influencing factors using random forest or LASSO regression algorithms.
7. A method for predicting the comfort level of liquor after drinking based on multi-stage tasting data, applied to the system according to any one of claims 1 to 6, characterized in that: The following steps are involved: (1) System initialization: perform self-tests on sensors and wearable devices, configure data sampling frequency, and clear historical cache; (2) Multi-stage data collection: Scan the code or manually enter the basic information of the liquor before drinking; During drinking, sensors monitor drinking behavior data in real time and record taster scores; After drinking, physiological data will be collected for ≥60 minutes and a user feedback questionnaire will be sent; (3) Data processing and analysis: Normalize key feature variables; The significant influencing factors are extracted through the feature screening algorithm to generate a feature matrix containing ≥10-dimensional features; (4) Prediction model operation: The feature matrix is input into the trained prediction model, and the quantitative score of post-drinking comfort and description of discomfort symptoms are output; (5) Result feedback and application: Real-time feedback of results to users to guide consumers’ drinking decisions, optimize tasters’ evaluation standards, and improve manufacturers’ processes.
8. The prediction method according to claim 7, wherein: Feature screening in step (3) includes: Random forest is used to calculate feature importance ranking; Redundant variables were eliminated through LASSO regression, and the factor combinations significantly correlated with post-drinking comfort were retained.
9. The prediction method according to claim 7, wherein: The output form of the prediction model in step (4) is: Comfort score: 0-10 point scale, with higher scores indicating better comfort; Description of discomfort symptoms: Output the probability of occurrence of headache, dry mouth, and gastrointestinal discomfort in the form of probability.
10. The prediction method according to claim 7, wherein: The manufacturer-side application in step (5) specifically includes: Based on the comfort rating of batch products after drinking, locate the threshold of wine components that cause discomfort; Adjust the distillation process or aging cycle based on the quantitative relationship between the combination of key factors and discomfort symptoms.