Weeping forsythia jasmine tea fragrance prediction and optimization system based on machine learning

Abstract

The invention discloses a fructus forsythiae jasmine tea aroma prediction and optimization system based on machine learning, and belongs to the technical field of tea processing and quality control, and the system comprises a raw material data acquisition module which is used for acquiring chemical components and other related data of fructus forsythiae tea in a preset batch, the other related data comprises appearance characteristics of the forsythia suspensa tea leaves of a preset batch and the production place and picking time of the forsythia suspensa tea leaves, and a database is constructed based on the raw material data acquisition module; and the characteristic acquisition module is used for acquiring the types and contents of aroma components of the jasmine flowers in a preset batch. According to the method, the comprehensive database is constructed by comprehensively considering the chemical components and the appearance characteristics of the fructus forsythiae tea and the aroma components and the processing technological parameters of the jasmine flowers, and the aroma prediction is performed by utilizing the machine learning model, so that the aroma characteristics of the fructus forsythiae-jasmine flower tea can be more accurately predicted, the aroma quality of the tea is favorably improved, and the economic benefit is increased. And meanwhile, quality fluctuation caused by process instability is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of tea processing and quality control, in particular to a forsythia jasmine tea aroma prediction and optimization system based on machine learning. Background Art

[0002] Forsythia jasmine tea, a scented tea with a unique flavor, is critically evaluated for its aroma quality. Traditionally, scented tea aroma evaluation relies primarily on manual sensory evaluation, a method characterized by subjectivity, inefficiency, and difficulty in quantification. Furthermore, numerous factors influence aroma formation during scented tea processing, including the quality of the tea leaves, the type and quality of the fresh flowers, and the scenting process parameters. These factors are difficult to precisely control for optimal aroma through experience and manual manipulation alone.

[0003] As a branch of artificial intelligence and computer science, machine learning is increasingly used in food quality prediction and optimization, which provides new ideas for solving the above problems.

[0004] However, the aroma quality of Forsythia jasmine tea is affected by a combination of factors, and the training of machine learning models requires a large amount of data to learn the complex relationship between features and aroma. However, in actual applications, due to limited production batches and high data collection costs, the amount of data available for training is insufficient, which affects the accuracy and generalization ability of the model, and thus cannot accurately predict its aroma quality. Summary of the Invention

[0005] In view of the above problems existing in the technical field of tea processing and quality control, the present invention is proposed.

[0006] Therefore, one of the objects of the present invention is to provide a machine learning-based aroma prediction and optimization system for Forsythia jasmine tea, which constructs a comprehensive database by comprehensively considering the chemical composition, appearance characteristics of Forsythia tea leaves, the aroma components of jasmine flowers, and processing parameters, and uses a machine learning model for aroma prediction. This method can more accurately predict the aroma characteristics of Forsythia jasmine tea, help improve the aroma quality of the flower tea, and reduce quality fluctuations caused by process instability.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] The present invention provides a machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea, comprising:

[0009] A raw material data acquisition module is used to obtain the chemical composition and other relevant data of a preset batch of Forsythia tea leaves, including the appearance characteristics, origin and picking time of the preset batch of Forsythia tea leaves, and to build a database based on the raw material data acquisition module;

[0010] a feature collection module for collecting the types and contents of aroma components of a preset batch of jasmine flowers, and also for collecting other features of the preset batch of jasmine flowers, such as variety, maturity, picking time, and processing method, to construct a database of the preset batch of jasmine flowers;

[0011] A processing parameter fusion processing module is used to obtain the process parameters of the preset batch of Forsythia suspensa jasmine tea; the processing parameter fusion processing module includes a division unit, an acquisition unit, an evaluation and collection unit, and an analysis unit;

[0012] The division unit is responsive to the raw material data acquisition module and is configured to divide the preset batches of Forsythia tea leaves into different groups, and further configured to divide the preset batches of Forsythia tea leaves into different groups according to appearance characteristics of the preset batches of Forsythia tea leaves;

[0013] The acquisition unit is responsive to the division of the preset batches of Forsythia tea leaves and is used to acquire process parameters of the different Forsythia tea leaves during the scenting process;

[0014] The evaluation collection unit responds to the acquisition unit and is used to perform aroma detection on different Forsythia tea leaves after scenting according to the process parameters to obtain aroma characteristics of different Forsythia tea leaves; and to perform aroma scoring according to the aroma characteristics;

[0015] The analyzing unit responds to the evaluation and collecting unit, and is used to analyze the aroma differences of different Forsythia suspensa tea leaves according to the aroma characteristics, and analyze the correlation effects between the aroma differences and different chemical components and / or appearance characteristics.

[0016] As a preferred solution of the present invention, wherein: in the raw material data acquisition module, the appearance characteristics include color and form, and the form includes branch form and flower form;

[0017] In the feature collection module, the processing methods include fresh-keeping processing, drying processing and storage processing.

[0018] As a preferred embodiment of the present invention, the division unit divides the preset batch of Forsythia tea leaves into categories according to the main chemical components, wherein the preset batch of Forsythia tea leaves are divided into tea polyphenols, alkaloids, amino acids, proteins, and sugars according to the main chemical components, and the proportion of different chemical components in each of the divided chemical components is calculated; based on the proportion, the preset batch of Forsythia tea leaves are divided into high-tea polyphenol-type Forsythia tea leaves, high-alkaloid-type Forsythia tea leaves, high-amino acid-type Forsythia tea leaves, high-protein-type Forsythia tea leaves, and / or high-sugar-type Forsythia tea leaves;

[0019] It also includes classification based on the sensory quality of chemical components, including dividing preset batches of Forsythia tea into light fragrance, floral fragrance, fruity fragrance and sweet fragrance types based on the sensory quality of chemical components.

[0020] As a preferred embodiment of the present invention, in the dividing unit, the preset batch of Forsythia tea leaves are divided according to the appearance characteristics of the preset batch of Forsythia tea leaves, and the division method includes dividing the preset batch of Forsythia tea leaves according to the leaf shape, leaf edge, leaf size, leaf color and leaf surface characteristics, wherein,

[0021] The leaf shapes include simple leaf type and compound leaf type;

[0022] The leaf edges include serrated and entire edges;

[0023] The leaf sizes include large leaf type and small leaf type;

[0024] The leaf colors include dark green and light yellow-green.

[0025] As a preferred solution of the present invention, in the acquisition unit, the process parameters include scenting temperature, scenting ambient humidity and scenting time, as well as stirring frequency, and form a data set about the process parameters.

[0026] As a preferred solution of the present invention, in the analysis unit, the analysis of the correlation impact includes the following steps:

[0027] The aroma characteristics are divided into ▽1, ▽2, ..., ▽ n , where ▽ n represents the nth aroma characteristic;

[0028] Calculate the proportion of different types of Forsythia tea leaves in different aroma characteristics;

[0029] Obtaining the process parameters of Forsythia suspensa tea corresponding to different aroma characteristics;

[0030] When making Forsythia jasmine tea in the future, if the same model of Forsythia tea and the process parameters corresponding to this model of Forsythia tea are selected for processing, the system will determine that the aroma characteristics produced are the same as the corresponding aroma characteristics divided; otherwise, no determination will be made.

[0031] As a preferred embodiment of the present invention, the chemical composition data of the preset batch of Forsythia tea leaves, the aroma component types and content data of jasmine flowers, as well as the process parameters and aroma scores are standardized, and the standardization includes cleaning, removal of outliers and missing values, and normalization. Feature data of different dimensions and numerical ranges are converted to a unified interval [0, 1], and a prediction model for aroma prediction is generated. At the same time, in the prediction model, the data is divided into a training set, a validation set, and a test set.

[0032] As a preferred solution of the present invention, in the prediction model, the chemical composition data of Forsythia suspensa tea leaves, the types and content data of aroma components of jasmine flowers, and the process parameters are used as input layer samples, and the aroma scores are used as output layer samples. The prediction model is trained using the data in the training set, and the hyperparameters in the prediction model are optimized and adjusted by cross-validation method to improve the generalization ability of the prediction model.

[0033] As a preferred solution of the present invention, before optimizing and adjusting the hyperparameters, the following steps are included:

[0034] According to the distinguished aroma characteristics, the proportion of the chemical components corresponding to each of the aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are obtained;

[0035] The obtained proportions of the chemical components corresponding to the respective aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are distinguished in a manner consistent with the manner in which the aroma characteristics are distinguished. The obtained proportions of the chemical components corresponding to the respective aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are distinguished as ▽ 1-1 ,▽ 2-1 ,...,▽ n-1 , where ▽ n-1 Indicates the proportion of a chemical component corresponding to the nth aroma characteristic, the type and content of the jasmine aroma component, and the process parameters;

[0036] The hyperparameter is lower and / or higher than the distinguished 1-1 、▽ 2-1 and ▽ n-1 The specific gravity of the corresponding chemical components, the types and contents of the aroma components of jasmine flowers and the process parameters;

[0037] Optimizing and adjusting the hyperparameters includes:

[0038] When making new Forsythia suspensa jasmine tea in the future, select the ones that are higher and / or lower than the ones that are distinguished. 1-1 、▽ 2-1 and ▽ n-1 The corresponding chemical composition ratio, jasmine aroma component type and content and process parameters are used to make it, and the values higher and / or lower than the distinguished values are recorded. 1-1 、▽ 2-1 and ▽ n-1 The corresponding proportions of chemical components, types and contents of jasmine aroma components and percentages of process parameters;

[0039] After the forsythia jasmine tea is prepared, the aroma characteristics of the forsythia jasmine tea are obtained and the new aroma characteristics are marked;

[0040] Optimization is performed according to the marked aroma characteristics, and the optimization method includes obtaining the proportion of chemical components corresponding to the aroma characteristics, the types and contents of the aroma components of jasmine flowers, and process parameters. On the basis of the corresponding proportion of chemical components, the types and contents of the aroma components of jasmine flowers, and process parameters, optimization is performed by adjusting the proportion of chemical components, the types and contents of the aroma components of jasmine flowers, and the process parameters by lowering them by 3% to 5% and / or increasing them by 3% to 5% to obtain new aroma characteristics.

[0041] Beneficial effects:

[0042] 1. The present invention can optimize the processing parameters, the proportion of chemical components, and the types and contents of jasmine aroma components based on the analysis results of the prediction model, thereby achieving precise control of the aroma of the scented tea, which helps to improve the aroma quality of the scented tea and reduce quality fluctuations caused by process instability;

[0043] 2. The present invention improves the generalization ability of machine learning by standardizing data and optimizing the hyperparameters in the prediction model using cross-validation, enabling the system to better adapt to tea and jasmine flowers from different batches and origins, and having wider applicability;

[0044] 3. During operation, the system can accumulate a large amount of data on tea chemical components, jasmine aroma components, processing parameters and aroma evaluation. These data can provide rich data resources and scientific basis for new product development, raw material procurement and process improvement of Forsythia jasmine tea, and will be beneficial for production companies to optimize the aroma of flower tea according to changes in the tastes of the market population. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0046] Figure 1 Schematic diagram of the modular structure of the machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of the process structure of an embodiment of the present invention;

[0048] Numbers in the figure: 110 - raw material data acquisition module; 120 - feature acquisition module; 130 - processing technology parameter fusion processing module; 1301 - division unit; 1302 - acquisition unit; 1303 - evaluation acquisition unit; 1304 - analysis unit. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.

[0050] Due to the limited production batches and high data collection costs of existing technologies, the amount of data available for training is insufficient, which affects the accuracy and generalization ability of the model, and makes it impossible to accurately predict its aroma quality.

[0051] Based on this, the present invention proposes a machine learning-based aroma prediction and optimization system for Forsythia jasmine tea. It constructs a comprehensive database by comprehensively considering the chemical composition, appearance characteristics of Forsythia tea leaves, the aroma components of jasmine flowers, and processing parameters, and uses a machine learning model for aroma prediction. This method can more accurately predict the aroma characteristics of Forsythia jasmine tea, help improve the aroma quality of the flower tea, and reduce quality fluctuations caused by process instability.

[0052] The present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0053] Reference Figures 1 to 2 , is an embodiment of the present invention, which provides a machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea, comprising:

[0054] A raw material data acquisition module 110 is used to acquire chemical components and other relevant data of a preset batch of Forsythia tea leaves, including the appearance characteristics, origin, and picking time of the preset batch of Forsythia tea leaves, and to construct a database based on the raw material data acquisition module;

[0055] In this embodiment, in a feasible implementation scheme, the chemical components (such as the content of tea polyphenols, amino acids, caffeine, etc.) of the Forsythia suspensa tea leaves are detected using a near-infrared spectrometer or other equipment to obtain spectral data of the Forsythia suspensa tea leaves, and then obtain their chemical components;

[0056] A feature collection module 120 is configured to collect the types and contents of aroma components of a preset batch of jasmine flowers, and also to collect other features of the preset batch of jasmine flowers, including variety, maturity, picking time, and processing method, to construct a database of the preset batch of jasmine flowers;

[0057] In this example, the raw material data acquisition module and feature collection module systematically collected multi-dimensional data such as the chemical composition, appearance characteristics, origin, picking time, variety, maturity, and processing methods of Forsythia suspensa tea and jasmine flowers, providing a rich information foundation for subsequent analysis and prediction.

[0058] Based on the collected data, a database of Forsythia tea and jasmine flowers was built to facilitate data storage, management, and access, providing reliable data support for subsequent machine learning model training and aroma prediction.

[0059] The processing parameter fusion processing module 130 is used to obtain the process parameters of a preset batch of Forsythia suspensa jasmine tea; the processing parameter fusion processing module 130 includes a division unit 1301, an acquisition unit 1302, an evaluation and collection unit 1303, and an analysis unit 1304;

[0060] The division unit 1301 responds to the raw material data acquisition module and is used to divide the preset batches of Forsythia tea leaves, and further includes a step of dividing the preset batches of Forsythia tea leaves according to the appearance characteristics of the preset batches of Forsythia tea leaves;

[0061] The acquisition module 1302 is responsive to the division of the preset batches of Forsythia tea leaves and is used to obtain the process parameters of the different Forsythia tea leaves during the scenting process;

[0062] The evaluation collection unit response acquisition unit 1303 is used to detect the aroma of different Forsythia tea leaves after scenting according to the process parameters, obtain the aroma characteristics of different Forsythia tea leaves; and perform aroma scoring based on the aroma characteristics;

[0063] In this embodiment, the aroma characteristics include aroma type, aroma intensity (concentration) and persistence;

[0064] In one feasible implementation, an electronic nose device is used to detect the aroma of processed forsythia jasmine tea to obtain its aroma characteristic fingerprint data. At the same time, a sensory evaluation team is organized to score and evaluate the aroma of the tea according to a standardized evaluation process, including indicators such as aroma type (such as light fragrance, sweet fragrance, floral fragrance, etc.), aroma intensity, and persistence. The electronic nose data and sensory evaluation data are combined to construct an aroma evaluation dataset.

[0065] The analysis unit 1304 responds to the evaluation collection unit and is used to analyze the aroma differences of different Forsythia tea leaves according to the aroma characteristics, and analyze the correlation between the aroma differences and different chemical components and / or appearance characteristics.

[0066] In the raw material data acquisition module, appearance features include color and shape, and shape includes branch shape and flower shape;

[0067] In this embodiment, the appearance characteristics of Forsythia suspensa tea leaves are further refined into color and morphology (including branch morphology and flower morphology), making the description of the appearance of the tea leaves more specific and accurate, which helps to more comprehensively evaluate the quality of the tea leaves;

[0068] In the feature collection module, the processing methods include fresh-keeping processing, drying processing and storage processing;

[0069] The processing methods of jasmine flowers were included in the feature collection scope, including fresh-keeping treatment, drying treatment and storage treatment. These processing methods have a significant impact on the aroma components and quality of jasmine flowers. Considering these factors can more accurately predict the aroma of scented tea.

[0070] In the classification unit, a predetermined batch of Forsythia tea leaves is classified according to the classification of main chemical components. According to the main chemical components, the predetermined batch of Forsythia tea leaves is classified into tea polyphenols, alkaloids, amino acids, proteins, and sugars. Among the classified chemical components, the proportion of different chemical components is calculated; based on the proportion, the predetermined batch of Forsythia tea leaves is classified into high tea polyphenol type Forsythia tea leaves, high alkaloid type Forsythia tea leaves, high amino acid type Forsythia tea leaves, high protein type Forsythia tea leaves, and / or high sugar type Forsythia tea leaves;

[0071] In this embodiment, tea polyphenols include catechins (such as epicatechin, epicatechin gallate, etc.), flavonoids (such as quercetin, kaempferol, etc.), anthocyanins, etc. Tea polyphenols are the most important antioxidant components in tea and have a significant impact on the color, aroma and taste of tea;

[0072] Alkaloids, mainly including caffeine, theobromine and theophylline, which give tea its bitter taste and also have certain physiological activities;

[0073] Amino acids, such as theanine, glutamic acid, and aspartic acid, are the main contributors to the freshness and refreshing taste of tea and are also involved in the formation of its aroma;

[0074] Protein: Tea contains a high amount of protein, but most of it is insoluble in water. Water-soluble protein has an important influence on the taste of tea.

[0075] Sugars, including monosaccharides (such as glucose and fructose), disaccharides (such as sucrose) and polysaccharides (such as cellulose, starch, and tea polysaccharides). Sugars not only affect the taste of tea, but also participate in the formation of its aroma and color.

[0076] It also includes classification based on the sensory quality of chemical components, including classifying preset batches of Forsythia tea into light fragrance, floral fragrance, fruity fragrance and sweet fragrance types according to the sensory quality of chemical components;

[0077] In this embodiment, the fragrance type: the volatile components contain high levels of low-boiling-point alcohols and aldehydes, and have a fresh aroma;

[0078] Floral fragrance: contains more high-boiling point alcohols and esters, such as phenylethyl alcohol and linalool, and has a floral fragrance;

[0079] Fruity type: contains more esters and aldehydes, with a fruity aroma;

[0080] Sweet aroma: high content of sugars and amino acids, with a sweet aroma;

[0081] In this embodiment, the Forsythia suspensa tea leaves are classified according to their main chemical components, and the proportion of each chemical component is calculated to further classify different types of tea leaves. This classification method helps to select the appropriate processing technology according to the chemical composition characteristics of the tea leaves to achieve the desired aroma quality;

[0082] Furthermore, in addition to the chemical composition classification, tea is also divided into light fragrance, floral fragrance, fruity fragrance, and sweet fragrance types based on the sensory quality of the chemical composition. Combining chemical composition with sensory quality is closer to consumers' actual experience of tea aroma, providing a more scientific and practical method for tea classification and quality evaluation.

[0083] In the division unit, the preset batch of Forsythia tea leaves are divided according to the appearance characteristics of the preset batch of Forsythia tea leaves, and the division method includes dividing the preset batch of Forsythia tea leaves according to the leaf shape, leaf edge, leaf size, leaf color and leaf surface characteristics, wherein,

[0084] Leaf shapes include simple and compound leaf types;

[0085] Simple leaf type: The leaves are single and have various shapes. Common ones include ovate, broad ovate, elliptical ovate, etc. For example, the leaves of common forsythia are usually ovate or broad ovate.

[0086] Compound leaf type: The leaves are 3-lobed or trifoliate, that is, one leaf is composed of multiple leaflets. This type of leaf is also common in Forsythia suspensa. It can be further distinguished by observing the degree of leaf division and the shape of the leaflets.

[0087] It also includes special shapes: For example, the leaves of Forsythia suspensa f.pubescens are soft or short-haired on the underside and have dense veins; while the leaves of Forsythia salikiangensis are glabrous on both sides and have entire margins;

[0088] Leaf margins include serrate and entire;

[0089] Serrated: The leaf margins have sharp or coarse serrations, which is a distinctive feature of Forsythia suspensa leaves. For example, the leaf margins of common Forsythia suspensa have sharp or coarse serrations except at the base.

[0090] Entire leaf margin: The leaf margin is smooth and without serrations. This type of leaf may appear in some special varieties, such as the entire leaf margin of Lijiang Forsythia suspensa.

[0091] Leaf sizes include large leaf type and small leaf type;

[0092] Large leaf type: The leaves are large, usually more than 5 cm long and more than 3 cm wide. For example, the leaves of common forsythia can be up to 10 cm long and 5 cm wide.

[0093] Small-leaf type: The leaves are small, less than 5 cm in length and less than 3 cm in width. For example, the leaves of Forsythia koreana are ovate or broadly ovate and relatively small.

[0094] Leaf colors include dark green and pale yellow-green;

[0095] Dark green: The upper side of the leaves is dark green, which is the common color of most forsythia leaves;

[0096] Light yellow-green: The underside of the leaf is light yellow-green. This color contrast helps to identify the front and back of the leaf.

[0097] Also includes special colors: some cultivated varieties may have special leaf colors, such as the golden-yellow leaves of Forsythia suspensa 'Aurea';

[0098] Leaf surface characteristics include smooth and hairy types;

[0099] Smooth type: The leaves are glabrous on both sides and have a smooth surface. For example, the leaves of Forsythia viridissima are glabrous on both sides.

[0100] Hairy type: The leaf surface has soft hairs or short hairs, for example, the underside of the leaf of Forsythia pubescens has soft hairs or short hairs;

[0101] In this embodiment, the classification method of the appearance characteristics of Forsythia suspensa tea leaves is further refined. This detailed classification helps to more accurately identify and distinguish different batches of tea leaves, providing a more specific basis for tea grading and processing.

[0102] In the acquisition unit, the process parameters include the scenting temperature, the scenting ambient humidity, the scenting time, and the stirring frequency, and a data set of the process parameters is formed;

[0103] In this embodiment, in a feasible implementation scheme, a temperature sensor is used to monitor the scenting temperature, a humidity sensor is used to obtain the scenting ambient humidity, and a time recording device (such as a counter) is used to record the scenting time;

[0104] In this example, the specific content of the processing parameters is clarified. These parameters are crucial to the aroma formation of Forsythia suspensa jasmine tea. Clarifying these parameters helps to more accurately control the processing process in subsequent analysis and optimization to achieve the desired aroma effect.

[0105] In the analysis unit, the analysis of the correlation impact includes the following steps:

[0106] Classify aroma features into ▽1, ▽2, ..., ▽ n , where ▽ n represents the nth aroma characteristic;

[0107] Calculate the proportion of different types of Forsythia tea leaves in different aroma characteristics;

[0108] Obtaining the process parameters of Forsythia suspensa tea corresponding to different aroma characteristics;

[0109] When making Forsythia jasmine tea in the future, if the same type of Forsythia tea leaves and the corresponding process parameters are selected for processing, the system will determine that the aroma characteristics produced are the same as the corresponding aroma characteristics classified; otherwise, no determination will be made;

[0110] In this example, Forsythia suspensa tea leaves with different aroma characteristics were analyzed, the proportion of different types of tea leaves in different aroma characteristics was calculated, and the corresponding process parameters were obtained. This analysis can reveal the intrinsic relationship between aroma differences and chemical composition, appearance characteristics and process parameters, providing a scientific basis for optimizing processing technology and improving aroma quality.

[0111] Based on the above, the chemical component data of the preset batches of Forsythia tea leaves, the aroma component types and content data of jasmine flowers, as well as the process parameters and aroma scores were standardized. The standardization process included cleaning, removing outliers and missing values, and normalization. The feature data of different dimensions and numerical ranges were converted to a unified interval of [0, 1]. A prediction model for aroma prediction was generated. At the same time, in the prediction model, the data was divided into training set, validation set, and test set.

[0112] In this embodiment, it can be used for subsequent training and tuning of the prediction model;

[0113] This processing method helps improve the quality and consistency of data, providing better data for training machine learning models, thereby improving the accuracy and generalization ability of the models;

[0114] Furthermore, in this embodiment, the prediction model uses the chemical component data of Forsythia suspensa tea leaves and the type and content data of the aroma components of jasmine flowers, as well as the process parameters, as input layer samples, and the aroma scores as output layer samples. The prediction model is trained using the data in the training set, and the hyperparameters in the prediction model are optimized and adjusted through cross-validation to improve the generalization ability of the prediction model.

[0115] In this embodiment, this training and optimization method can improve the model's adaptability to different data and further enhance the model's prediction accuracy and generalization ability;

[0116] Before optimizing the hyperparameters, the following steps are included:

[0117] According to the distinguished aroma characteristics, the proportion of chemical components corresponding to each aroma characteristic, the types and contents of the aroma components of jasmine flowers, and process parameters are obtained;

[0118] The obtained proportions of chemical components corresponding to each aroma characteristic, the types and contents of jasmine aroma components, and process parameters are distinguished in the same way as the aroma characteristics are distinguished. The obtained proportions of chemical components corresponding to each aroma characteristic, the types and contents of jasmine aroma components, and process parameters are distinguished as ▽ 1-1 ,▽ 2-1 ,...,▽ n-1 , where ▽ n-1 Indicates the proportion of a chemical component corresponding to the nth aroma characteristic, the type and content of the jasmine aroma component, and the process parameters;

[0119] Hyperparameters are lower and / or higher than the distinguished ▽ 1-1 、▽ 2-1 and ▽ n-1The specific gravity of the corresponding chemical components, the types and contents of the aroma components of jasmine flowers and the process parameters;

[0120] The optimization and adjustment of hyperparameters include:

[0121] When making new Forsythia suspensa jasmine tea in the future, select the ones that are higher and / or lower than the ones that are distinguished. 1-1 、▽ 2-1 and ▽ n-1 The corresponding chemical composition ratio, jasmine aroma component type and content and process parameters are used to make it, and the values higher and / or lower than the distinguished values are recorded. 1-1 、▽ 2-1 and ▽ n-1 The corresponding proportions of chemical components, types and contents of jasmine aroma components and percentages of process parameters;

[0122] After the forsythia jasmine tea is prepared, the aroma characteristics of the forsythia jasmine tea are obtained and the new aroma characteristics are marked;

[0123] Optimizing according to the marked aroma characteristics, the optimization method includes obtaining the specific gravity of the chemical components, the types and contents of the aroma components of jasmine flowers, and process parameters corresponding to the aroma characteristics, and optimizing by adjusting the specific gravity of the chemical components, the types and contents of the aroma components of jasmine flowers, and the process parameters by decreasing by 3% to 5% and / or increasing by 3% to 5% to obtain a new aroma characteristic;

[0124] In this embodiment, the adjustment of the process parameters includes adjusting the scenting temperature, the scenting ambient humidity, the scenting time, and the frequency of stirring the Forsythia suspensa tea leaves and jasmine flowers;

[0125] Before optimizing hyperparameters, we first obtain the chemical component ratios, aroma component types and contents, and process parameters corresponding to different aroma characteristics, and then differentiate them according to aroma characteristics. This optimization method can adjust the chemical component ratios, aroma components, and process parameters in a targeted manner according to specific aroma characteristic requirements, thereby obtaining aroma characteristics that better meet the expectations.

[0126] In the future, when making new Forsythia jasmine tea, it will be made according to the optimized parameters and the relevant data will be recorded. Then, it will be marked and further optimized according to the actual aroma characteristics of the flower tea produced. This dynamic optimization process can continuously adjust and improve the processing technology to adapt to different raw materials and market demands, and continuously improve the aroma quality of the flower tea.

[0127] In summary, the present invention constructs a comprehensive database by comprehensively considering the chemical composition and appearance characteristics of Forsythia tea leaves, the aroma components of jasmine flowers, and processing parameters, and uses a machine learning model for aroma prediction. This method can more accurately predict the aroma characteristics of Forsythia jasmine tea, help improve the aroma quality of flower tea, and reduce quality fluctuations caused by process instability.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea, characterized by: include: A raw material data acquisition module is used to obtain the chemical composition and other relevant data of a preset batch of Forsythia tea leaves, including the appearance characteristics, origin and picking time of the preset batch of Forsythia tea leaves, and to build a database based on the raw material data acquisition module; a feature collection module for collecting the types and contents of aroma components of a preset batch of jasmine flowers, and also for collecting other features of the preset batch of jasmine flowers, such as variety, maturity, picking time, and processing method, to construct a database of the preset batch of jasmine flowers; A processing parameter fusion processing module is used to obtain the process parameters of the preset batch of Forsythia suspensa jasmine tea; the processing parameter fusion processing module includes a division unit, an acquisition unit, an evaluation and collection unit, and an analysis unit; The division unit is responsive to the raw material data acquisition module and is configured to divide the preset batches of Forsythia tea leaves into different groups, and further configured to divide the preset batches of Forsythia tea leaves into different groups according to appearance characteristics of the preset batches of Forsythia tea leaves; The acquisition unit is responsive to the division of the preset batches of Forsythia tea leaves and is used to acquire process parameters of the different Forsythia tea leaves during the scenting process; The evaluation and collection unit responds to the acquisition unit and is used to perform aroma detection on different Forsythia tea leaves after scenting according to the process parameters to obtain aroma characteristics of different Forsythia tea leaves; and performing aroma scoring according to the aroma characteristics; The analyzing unit responds to the evaluation and collecting unit, and is used to analyze the aroma differences of different Forsythia suspensa tea leaves according to the aroma characteristics, and analyze the correlation effects between the aroma differences and different chemical components and / or appearance characteristics.

2. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 1, characterized in that: In the raw material data acquisition module, the appearance features include color and shape, and the shape includes branch shape and flower shape; In the feature collection module, the processing methods include fresh-keeping processing, drying processing and storage processing.

3. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 1, characterized in that: In the classification unit, a preset batch of Forsythia tea leaves is classified according to the classification of main chemical components, wherein the preset batch of Forsythia tea leaves is classified into tea polyphenols, alkaloids, amino acids, proteins and sugars according to the main chemical components, and the proportion of different chemical components in each of the classified chemical components is calculated; based on the proportions, the preset batch of Forsythia tea leaves is classified into high tea polyphenol type Forsythia tea leaves, high alkaloid type Forsythia tea leaves, high amino acid type Forsythia tea leaves, high protein type Forsythia tea leaves and / or high sugar type Forsythia tea leaves; It also includes classification based on the sensory quality of chemical components, including dividing preset batches of Forsythia tea into light fragrance, floral fragrance, fruity fragrance and sweet fragrance types based on the sensory quality of chemical components.

4. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 3, characterized in that: In the division unit, the preset batch of Forsythia tea leaves are divided according to the appearance characteristics of the preset batch of Forsythia tea leaves, and the division method includes dividing the preset batch of Forsythia tea leaves according to the leaf shape, leaf edge, leaf size, leaf color and leaf surface characteristics, wherein, The leaf shapes include simple leaf type and compound leaf type; The leaf edges include serrated and entire edges; The leaf sizes include large leaf type and small leaf type; The leaf colors include dark green and light yellow-green.

5. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 1, characterized in that: In the acquisition unit, the process parameters include scenting temperature, scenting ambient humidity and scenting time, as well as stirring frequency, and form a data set about the process parameters.

6. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 1, characterized in that: In the analysis unit, the analysis of the correlation impact includes the following steps: The aroma characteristics are divided into ▽1, ▽2, ..., ▽ n , where ▽ n represents the nth aroma characteristic; Calculate the proportion of different types of Forsythia tea leaves in different aroma characteristics; Obtaining the process parameters of Forsythia suspensa tea corresponding to different aroma characteristics; When making Forsythia jasmine tea in the future, if the same model of Forsythia tea and the process parameters corresponding to this model of Forsythia tea are selected for processing, the system will determine that the aroma characteristics produced are the same as the corresponding aroma characteristics divided; otherwise, no determination will be made.

7. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 1, characterized in that: The chemical composition data of the preset batch of Forsythia tea leaves, the type and content data of the aroma components of jasmine flowers, as well as the process parameters and aroma scores are standardized. The standardization includes cleaning, removing outliers and missing values, and normalization. Feature data of different dimensions and numerical ranges are converted to a unified interval [0, 1], and a prediction model for aroma prediction is generated. At the same time, in the prediction model, the data is divided into a training set, a validation set, and a test set.

8. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 7, characterized in that: In the prediction model, the chemical composition data of Forsythia suspensa tea leaves, the types and content data of jasmine aroma components, and the process parameters are used as input layer samples, and the aroma scores are used as output layer samples. The prediction model is trained using the data in the training set, and the hyperparameters in the prediction model are optimized and adjusted through cross-validation to improve the generalization ability of the prediction model.

9. The machine learning-based aroma prediction and optimization system for Forsythia suspensa jasmine tea according to claim 8, characterized in that: Before optimizing hyperparameters, include: According to the distinguished aroma characteristics, the proportion of the chemical components corresponding to each of the aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are obtained; The obtained proportions of the chemical components corresponding to the respective aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are distinguished in a manner consistent with the manner in which the aroma characteristics are distinguished. The obtained proportions of the chemical components corresponding to the respective aroma characteristics, the types and contents of the aroma components of jasmine flowers, and the process parameters are distinguished as ▽ 1-1 ,▽ 2-1 ,...,▽ n-1 , where ▽ n-1 Indicates the proportion of a chemical component corresponding to the nth aroma characteristic, the type and content of the jasmine aroma component, and the process parameters; The hyperparameter is lower and / or higher than the distinguished 1-1 、▽ 2-1 and ▽ n-1 The specific gravity of the corresponding chemical components, the types and contents of the aroma components of jasmine flowers and the process parameters; Optimizing and adjusting the hyperparameters includes: When making new Forsythia suspensa jasmine tea in the future, select the ones that are higher and / or lower than the ones that are distinguished. 1-1 、▽ 2-1 and ▽ n-1 The corresponding chemical composition ratio, jasmine aroma component type and content and process parameters are used to make it, and the values higher and / or lower than the distinguished values are recorded. 1-1 、▽ 2-1 and ▽ n-1 The corresponding proportions of chemical components, types and contents of jasmine aroma components and percentages of process parameters; After the forsythia jasmine tea is prepared, the aroma characteristics of the forsythia jasmine tea are obtained and the new aroma characteristics are marked; Optimization is performed according to the marked aroma characteristics, and the optimization method includes obtaining the proportion of chemical components corresponding to the aroma characteristics, the types and contents of the aroma components of jasmine flowers, and process parameters. On the basis of the corresponding proportion of chemical components, the types and contents of the aroma components of jasmine flowers, and process parameters, optimization is performed by adjusting the proportion of chemical components, the types and contents of the aroma components of jasmine flowers, and the process parameters by lowering them by 3% to 5% and / or increasing them by 3% to 5% to obtain new aroma characteristics.

Images