Intelligent automatic coffee machine based on big data analysis

Abstract

The invention discloses an intelligent automatic coffee machine based on big data analysis, and relates to the technical field of intelligent coffee makers, the intelligent automatic coffee machine comprises a coffee machine and an intelligent control system connected with the coffee machine, and the coffee machine comprises a conveying device and a heat preservation cabinet device; the intelligent control system comprises an information acquisition unit, a data analysis unit, a control unit and an APP matched with the control unit, the information acquisition unit is used for acquiring related information and sending the related information to the data analysis unit, and the data analysis unit is used for analyzing, processing and learning the received information and feeding back an analysis result to the control unit; the system has the advantages that the use convenience and satisfaction degree of the user are remarkably improved through remote ordering, diversified selection and a flexible goods taking mode, the system can provide accurate coffee formula recommendation based on user historical data and environmental factors, personalized requirements are met, and recommendation is more intelligent and accurate through advanced recommendation algorithm and model optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent coffee machines, and more specifically, to an intelligent automatic coffee machine based on big data analysis. Background Art

[0002] With the advancement of technology and the improvement of living standards, coffee has become one of the indispensable drinks in people's daily lives. In order to meet the growing personalized needs of consumers, smart coffee machines have gradually entered the market and become the new favorite of modern homes and commercial places. Smart coffee machines not only improve the convenience and safety of coffee making, but also realize the personalized customization of coffee taste through advanced scientific and technological means.

[0003] Smart coffee machines on the market now have some advanced features, such as adjusting the strength, temperature and taste of coffee according to the user's preferences and habits. These machines learn the user's drinking habits through built-in intelligent algorithms and automatically adjust the coffee making parameters to provide coffee that better suits personal tastes. In addition, smart coffee machines are also equipped with automatic cleaning functions and regular maintenance reminders, further improving the convenience of use and the durability of the equipment.

[0004] For example, a coffee machine based on an artificial intelligence algorithm disclosed in Chinese invention application number CN202111652019.0 uses a mathematical model obtained through artificial neural network learning and training, and takes the input coffee parameter data as the input layer to control the water-powder ratio and brewing temperature to complete the coffee making process, providing users with a direct and easy-to-use personalized coffee customization experience.

[0005] However, although the existing smart coffee system meets the personalized needs of users to a certain extent, there are still many links to be optimized in the entire process of coffee making, sales and pickup. For example, when consumers choose coffee types and adjust production parameters, they often need to go to the coffee machine in person to operate, lacking the convenience of remote control; after the coffee is made, the pickup method is also relatively simple, which cannot meet the needs of different users for immediate pickup, thermal storage or delayed pickup.

[0006] Therefore, in order to solve the above technical problems, the present invention proposes an intelligent automatic coffee machine based on big data analysis. Summary of the invention

[0007] In view of the deficiencies in the prior art, an object of the present invention is to provide an intelligent automatic coffee machine based on big data analysis.

[0008] To achieve the above object, the present invention provides the following technical solution: an intelligent automatic coffee machine based on big data analysis, comprising:

[0009] A coffee machine and an intelligent control system connected to the coffee machine, wherein the coffee machine includes a conveying device and a warming cabinet device;

[0010] The intelligent control system includes an information acquisition unit, a data analysis unit, a control unit, and an APP that matches the control unit;

[0011] The information acquisition unit is used to acquire relevant information and send it to the data analysis unit;

[0012] The data analysis unit is used to analyze, process and learn the received information, and feed back the analysis results to the control unit;

[0013] The control unit is used to control the operation of the coffee machine;

[0014] The APP is used to receive information from the data analysis unit, information acquisition unit and control unit, and feed it back to the user. At the same time, the coffee machine can be remotely controlled through the control unit.

[0015] Furthermore, the information acquisition unit can collect the user's historical coffee selection records and obtain real-time external environmental data such as the weather, season, time period, etc. of the current location, and can also obtain sales status and user feedback of different coffee recipes.

[0016] Furthermore, the data analysis unit analyzes the data including the following processing steps:

[0017] S1: preprocess the data, identify and fill missing values, and remove abnormal data to obtain a data set; S2: build personal portraits for different users based on the data set obtained in step S1, record their preferences, historical choices and other related data, and establish a relevant database;

[0018] S3: predicting preferences of different individual users based on their personal portraits in step S2, calculating scores of different coffee recipes based on the user's preferences, and calculating similarities of the coffee scores;

[0019] S4: Use the generative recommendation algorithm to find the recipe group that is most similar to the current coffee recipe and use it as the corresponding weight calculation in the data set;

[0020] S5: predict the user's ratings for different recipes, and select the recipe with the highest rating to recommend to the user;

[0021] S6: Extract the characteristic parameters of each coffee recipe and represent them as a feature vector. Then, build a user preference model based on the user's historical selections, and generate recommendations using the same similarity calculation and recommendation algorithm as in step S1.

[0022] S7: A hybrid recommendation algorithm is used to comprehensively consider user similarity and coffee recipe feature similarity, combine the comprehensive user portrait algorithm with the coffee recipe feature highlighting algorithm, and combine collaborative filtering and content-based recommendation algorithms to provide more accurate and personalized recommendations;

[0023] S8: Extract useful features from the data set obtained in step S1, including user preference features, coffee recipe features, and environmental features, and use them as feature training data sets.

[0024] Furthermore, the similarity calculation method of coffee ratings in step S2 is:

[0025]

[0026] Among them, r ui and r vi They represent the ratings of user u and user v on coffee recipe i respectively.

[0027] Furthermore, according to the hybrid recommendation algorithm in step S7, a weighted average method is used to calculate:

[0028]

[0029] in, It is the evaluation part of collaborative filtering. It is the content-based evaluation part, and α is the weight parameter, which is used to adjust the weights of the two algorithms to obtain the best recommendation solution.

[0030] Furthermore, in step S8, the data set is divided into a training set, a validation set, and a test set in a ratio of 70%:15%:15%. The gradient descent method is used to optimize the model parameters, and the early stopping strategy is adopted to prevent overfitting. Subsequently, the accuracy, recall, F1-score, and root mean square error (RMSE) are used to evaluate the model performance. The SVD algorithm formula is: R≈UCV T ;

[0031] Among them, R is the user-recipe rating matrix, U and V are low-dimensional feature matrices, C is a diagonal matrix, and the GMF model is:

[0032]

[0033] in, is the predicted score, p u and q i are the latent factor vectors for users and recipes, respectively.

[0034] Furthermore, hyperparameter optimization is performed based on the above model:

[0035] Adjust the learning rate of the model to balance the training speed and convergence; introduce L2 regularization to adjust the regularization strength; adjust the number of latent factors in the matrix decomposition to balance the model complexity and prediction performance;

[0036] The grid search algorithm and the random search algorithm are used to evaluate the hyperparameter combination and select the initial optimal hyperparameter combination; then the Bayesian optimization algorithm is used to iteratively select the optimal hyperparameter combination based on the Bayesian statistical model. The L2 regularization formula is:

[0037]

[0038] Where L is the loss function, λ is the regularization parameter, and P and Q are the latent factor matrices of users and recipes, respectively.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. In the present invention, a coffee machine system integrating intelligent production, sales, sorting and storage is proposed. The system not only supports users to remotely select coffee recipes and adjust production parameters through software and applets, but also provides a variety of pickup methods, including instant pickup, insulated storage pickup and delayed pickup, to meet the needs of different users. In addition, the system also integrates an intelligent recommendation system, which provides users with accurate and personalized coffee recipe suggestions through collaborative filtering and content-based recommendation algorithms, further improving user experience and service efficiency. DETAILED DESCRIPTION

[0041] The present invention provides an intelligent automatic coffee machine based on big data analysis, comprising a coffee machine and an intelligent control system connected to the coffee machine, wherein the coffee machine comprises a conveying device and a heat preservation cabinet device;

[0042] The intelligent control system includes an information acquisition unit, a data analysis unit, a control unit, and an APP that matches the control unit;

[0043] The information acquisition unit is used to acquire relevant information and send it to the data analysis unit;

[0044] The data analysis unit is used to analyze and process the received information and feed back the analysis results to the control unit;

[0045] The control unit is used to control the operation of the coffee machine;

[0046] The APP is used to receive information from the data analysis unit, information acquisition unit and control unit, and feed it back to the user. At the same time, the control unit can be remotely controlled;

[0047] In summary, compared with the current problem that users need to wait in line for a long time to select coffee, and are entangled in the diversity of coffee types when purchasing coffee, the present invention presets a variety of coffee beans, coffee companion additives, etc. in the coffee machine in advance, and users can purchase them online in advance through the APP, and adjust relevant parameters such as the production temperature, ice amount, sweetness, etc.

[0048] At the same time, after analysis by the data analysis unit, the user can be provided with recommended coffee types and related parameters, which can save the user's decision-making time and thus improve the user experience;

[0049] The information acquisition unit can collect the user's historical coffee selection records and obtain real-time external environmental data such as the current location's weather, season, time period, etc. It can also obtain sales and user feedback on different coffee recipes;

[0050] The data analysis unit analyzes the data according to the above information acquired by the information acquisition unit, including the following steps:

[0051] Preprocess the data, use the Aver algorithm to identify and fill missing values ​​in the acquired data to ensure data integrity, identify outliers in the data through statistical analysis and business rules, and use the Z-Score (standard score method) and IQR (interquartile range method) algorithms for the above data sets to filter out special and abnormal data that do not have common reference value in the data sets, perform data correction on abnormal data according to business rules, and consider eliminating values ​​that are too abnormal;

[0052] Get the weight value of each field in the data set, count and calculate the weight standard of each field, and enter the next step if the value is within the preset range. If it exceeds the threshold, it will enter the outlier set. IQR is the same as Z-Score, and its interquartile range is calculated. For values ​​lower than Q1-Q1.5IQR or higher than Q3+Q1.5IQR, they are considered outliers;

[0053] For the data processed in the above steps, the Dup-H algorithm is used to merge the fields with the same weight and value, and the frequency assignment calculation module in Dup-H is used to update the weight of the merged fields;

[0054] The data fields are normalized with the weight value as the primary key, and the weight values ​​are divided and stepped into the range of [0,1] using MAX-MIN. The Z-ScoreNor algorithm is used to convert the data to obtain a valuable data set.

[0055] Based on the data set obtained in the above steps, build personal portraits for different users and record their preferences, historical choices and other related data. At the same time, build corresponding databases for the environmental data, coffee order data (coffee recipe composition, production parameters and user evaluation) and sales data (coffee sales of different coffee recipes, user feedback and popularity index) in the data set;

[0056] Predict the preferences of different individual users based on their personal profiles in the above steps, and calculate the scores of different coffee recipes analyzed based on the preferences of the users;

[0057] First, we construct a U-RRM (user-recipe rating matrix), where each element in the matrix represents a user's rating of a coffee recipe.

[0058] Then, the similarity of coffee scores was calculated by cross-averaging using the ConsS (cosine similarity) and PearCor (Pearson correlation system) methods. The calculation formula is as follows:

[0059]

[0060] (where r ui and r vi Respectively represent the ratings of user u and user v on coffee recipe i)

[0061] Subsequently, the RecomGen algorithm (recommendation generation algorithm) is used to find the recipe group that is most similar to the current coffee recipe, and based on the scores of these recipes, the MatrixDPO (matrix distributed array computing) calculation method is used to calculate the corresponding weights of the data sets in the above data to obtain the final scoring matrix, and recommendations are generated based on this matrix.

[0062]

[0063] Predict user u’s rating for recipe i

[0064] Where N(i) represents the K recipes that are closest to recipe i, sim(ij) is the similarity between recipes i and j, and r uj is the rating of user u on recipe j.

[0065] At this point, we can predict a user's Mocha (referring to a coffee recipe or coffee type) rating for multiple recipes, and select the highest-rated recipe to recommend to the user;

[0066] Secondly, Count-BasedFilter (content-based recommendation algorithm) is used to assist in recommendation. The main steps are: first, extract the characteristic parameters of each coffee recipe and represent them as a feature vector; then, build a user preference model based on the user's historical choices (the user preference model is a weighted average of the user's historical choice weights);

[0067] Finally, the recommendation is generated using the similarity calculation and recommendation algorithm described above;

[0068] For the above two algorithms, the Hybrid-RecomSys algorithm is adopted. By comprehensively considering the user similarity and the feature similarity of the coffee recipe, algorithm 1 integrates the user portrait, and algorithm 2 highlights the characteristics of the coffee recipe. It combines collaborative filtering and content-based recommendation algorithms to provide more accurate and personalized recommendations.

[0069] The recommendation matrix generated by the above two algorithms is converted into a K-nearest neighbor list, and the Hybird-RecomSys strategy is used to fuse the recommendation results, using weighted average:

[0070]

[0071] in, It is the evaluation part of collaborative filtering. It is the content-based evaluation part, and α is the weight parameter used to adjust the weights of the two algorithms;

[0072] Sort by fusion scoring results to get the best recommendation solution;

[0073] Extract useful features from the data set obtained in the above steps, including user preference features, coffee recipe features, and environmental features, and use them as feature training data sets;

[0074] Select the SVD (Singular Value Decomposition) algorithm based on Matrix Factorization and use the Generalized Matrix Decomposition (GMF) model to train the data set;

[0075] First, the dataset is divided into training set, validation set, and test set in a ratio of 70%:15%:15%.

[0076] Secondly, the gradient descent method is used to optimize the model parameters, and the early stopping strategy is adopted to prevent overfitting;

[0077] Subsequently, the model performance was evaluated using indicators such as precision, recall, F1-Score, and root mean square error (RMSE);

[0078] SVD algorithm formula: R≈UCV T ;

[0079] Among them, R is the user-recipe rating matrix, U and V are low-dimensional feature matrices, and C is a diagonal matrix.

[0080] GMF Model:

[0081]

[0082] in, is the predicted score, p u and q i are the latent factor vectors for users and recipes respectively

[0083] Hyperparameter optimization is performed based on the above model.

[0084] (1) Adjust the learning rate of the model to balance the training speed and convergence; introduce L2 regularization to adjust the regularization strength; adjust the number of latent factors in the matrix decomposition to balance the model complexity and prediction performance;

[0085] (2) Use the GS (grid search) algorithm and RS (random search) algorithm to evaluate the hyperparameter combination and select the initial optimal hyperparameter combination; then use the BO (Bayesian optimization) algorithm to iteratively select the optimal hyperparameter combination based on the Bayesian statistical model;

[0086] L2 regularization formula:

[0087]

[0088] Where L is the loss function, λ is the regularization parameter, P and Q are the latent factor matrices of users and recipes, respectively;

[0089] The prediction results of multiple base models are used as new features to train the second-layer model for final prediction;

[0090] At the same time, combined with online learning methods, as new data arrives, model parameters are updated in real time to ensure that the model can adapt to changes in user behavior and conduct continuous learning and optimization;

[0091] In terms of big data processing, Hadoop HDFS is used to solve storage problems, and the Flink big data processing framework is used for data cleaning, feature engineering, and model training;

[0092] In terms of distributed computing, we use the PyTorch deep learning framework for training;

[0093] After the user logs in to the APP, the APP page will display columns such as system recommendations, my preferences, and past orders;

[0094] Among them, in the submenu under System Recommendation, there are multiple coffee types recommended by the system analysis. Clicking it will display the detailed specific types (coffee bean type, coffee companion type, sugar amount, temperature, ice amount, etc.);

[0095] Among them, the submenu under My Favorites displays the coffee information that users have set in the app in the past, such as Favorites 1, 2, and 3. Users can click on a specific coffee type and adjust the parameters on the page (including coffee bean type, coffee companion type, sugar amount, temperature, ice amount, etc.), and choose whether to overwrite the favorite coffee;

[0096] Among them, the past order information is displayed in the submenu under the past orders. Users can click into the historical orders and directly enter the purchase order from the historical orders. The selected historical order parameters are automatically imported into the new order information and can be adjusted;

[0097] After the user has selected the coffee type, the system analyzes the coffee making time and adds the making time to the current time, which is the stand-up time. After the stand-up time, the system provides the user with a time option for picking up the coffee. If the coffee picking time selected by the user is greater than the stand-up time, the system pops up two options for the user to choose from: one is to make coffee immediately, and after the coffee is made, the coffee is stored in the thermal cabinet device through the conveyor device; the other is to make coffee according to the estimated coffee picking time. If there are many orders, the system will automatically make coffee for the users who placed the orders later in advance according to the order before and after, and store it in the thermal cabinet device, and prompt the user through the APP;

[0098] It is worth noting that the coffee order information will be updated in real time within the APP system, which is convenient for users to observe in real time;

[0099] Through remote ordering, diversified choices and flexible pickup methods, the user's convenience and satisfaction have been significantly improved. The system can provide accurate coffee recipe recommendations based on user historical data and environmental factors to meet personalized needs. It uses advanced recommendation algorithms and model optimization to make recommendations more intelligent and accurate, realizes automatic sorting and storage of coffee, improves efficiency, reduces manual intervention, provides an intuitive APP interface, supports remote control and real-time monitoring, and APP pushes timely notifications to users, improving user experience.

[0100] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any ordinary technician in the industry can smoothly implement the present invention as described above. However, any equivalent changes, modifications and evolutions made by technicians familiar with the profession without departing from the scope of the technical solution of the present invention using the technical content disclosed above are all equivalent embodiments of the present invention. At the same time, any equivalent changes, modifications and evolutions made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent automatic coffee machine based on big data analysis, characterized in that: include: A coffee machine and an intelligent control system connected to the coffee machine, wherein the coffee machine includes a conveying device and a warming cabinet device; The intelligent control system includes an information acquisition unit, a data analysis unit, a control unit, and an APP that matches the control unit; The information acquisition unit is used to acquire relevant information and send it to the data analysis unit; The data analysis unit is used to analyze, process and learn the received information, and feed back the analysis results to the control unit; The control unit is used to control the operation of the coffee machine; The APP is used to receive information from the data analysis unit, information acquisition unit and control unit, and feed it back to the user. At the same time, the coffee machine can be remotely controlled through the control unit.

2. The intelligent automatic coffee machine based on big data analysis according to claim 1, characterized in that: The information acquisition unit can collect the user's historical coffee selection records and obtain real-time external environmental data such as the weather, season, time period, etc. of the current location, and can also obtain sales conditions and user feedback of different coffee recipes.

3. The intelligent automatic coffee machine based on big data analysis according to claim 2, characterized in that: The data analysis unit analyzes the data including the following processing steps: S1: Preprocess the data, identify and fill missing values, and remove abnormal data to obtain the data set; S2: Based on the data set obtained in step S1, construct personal portraits for different users, record their preferences, historical choices and other related data, and establish a relevant database; S3: predicting preferences of different individual users based on their personal portraits in step S2, calculating scores of different coffee recipes based on the user's preferences, and calculating similarities of the coffee scores; S4: Use the generative recommendation algorithm to find the recipe group that is most similar to the current coffee recipe and use it as the corresponding weight calculation in the data set; S5: predict the user's ratings for different recipes, and select the recipe with the highest rating to recommend to the user; S6: Extract the characteristic parameters of each coffee recipe and represent them as a feature vector. Then, build a user preference model based on the user's historical selections, and generate recommendations using the same similarity calculation and recommendation algorithm as in step S1. S7: A hybrid recommendation algorithm is used to comprehensively consider user similarity and coffee recipe feature similarity, combine the comprehensive user portrait algorithm with the coffee recipe feature highlighting algorithm, and combine collaborative filtering and content-based recommendation algorithms to provide more accurate and personalized recommendations; S8: Extract useful features from the data set obtained in step S1, including user preference features, coffee recipe features, and environmental features, and use them as feature training data sets.

4. The intelligent automatic coffee machine based on big data analysis according to claim 3, characterized in that: The similarity calculation method of coffee ratings in step S2 is: Among them, r ui and r vi They represent the ratings of user u and user v on coffee recipe i respectively.

5. The intelligent automatic coffee machine based on big data analysis according to claim 4, characterized in that: In step S5, predict the rating of user u for recipe i Where N(i) represents the k recipes closest to recipe i, sim(ij) is the similarity between recipes i and j, and r ui is the rating of user u on recipe j.

6. The intelligent automatic coffee machine based on big data analysis according to claim 5, characterized in that: According to the hybrid recommendation algorithm in step S7, the weighted average method is used to calculate: in, It is the evaluation part of collaborative filtering. It is the content-based evaluation part, and α is the weight parameter, which is used to adjust the weights of the two algorithms to obtain the best recommendation solution.

7. The intelligent automatic coffee machine based on big data analysis according to claim 6, characterized in that: In step S8, the data set is divided into a training set, a validation set, and a test set in a ratio of 70%:15%:15%. The gradient descent method is used to optimize the model parameters, and the early stopping strategy is adopted to prevent overfitting. Subsequently, the accuracy, recall, F1-score, and root mean square error (RMSE) are used to evaluate the model performance. The SVD algorithm formula is: R≈UCV T ; Among them, R is the user-recipe rating matrix, U and V are low-dimensional feature matrices, C is a diagonal matrix, and the GMF model is: in, is the predicted score, p u and q i are the latent factor vectors for users and recipes, respectively.

8. The intelligent automatic coffee machine based on big data analysis according to claim 7, characterized in that: Perform hyperparameter optimization based on the above model: Adjust the learning rate of the model to balance the training speed and convergence; introduce L2 regularization to adjust the regularization strength; adjust the number of latent factors in the matrix decomposition to balance the model complexity and prediction performance; The grid search algorithm and the random search algorithm are used to evaluate the hyperparameter combination and select the initial optimal hyperparameter combination; then the Bayesian optimization algorithm is used to iteratively select the optimal hyperparameter combination based on the Bayesian statistical model. The L2 regularization formula is: Where L is the loss function, λ is the regularization parameter, and P and Q are the latent factor matrices of users and recipes, respectively.

Images