Cosmetic Adverse Reaction Analysis and Feedback Method and System Based on Active Monitoring

Abstract

The present invention discloses a method and system for analyzing and feedbacking adverse reactions to cosmetics based on active monitoring, relating to the technical field of cosmetics information management. This method, based on active monitoring, obtains usage monitoring feedback information for cosmetics across different age groups and performs predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group. Furthermore, the system analyzes and obtains a skin adverse feedback index and a usage adverse feedback index for each age group. Furthermore, the system obtains a cosmetic ingredient feedback safety index and combines these indices for analysis to obtain a comprehensive adverse feedback index for each age group. The system compares the comprehensive adverse feedback index with a preset comprehensive adverse feedback index threshold, generates feedback suggestions, and labels them. This effectively captures immediate feedback from users of different age groups during the use of cosmetics, providing targeted analysis and subsequently accurately assessing cosmetic adverse reactions.

Description

Technical Field

[0001] The present invention relates to the technical field of cosmetics information management, and in particular to a method and system for analyzing and feeding back cosmetics adverse reactions based on active monitoring. Background Art

[0002] As consumers pay more and more attention to the quality and effects of cosmetics, the safety and applicability of cosmetics have become core issues in production and marketing. Consumers' skin reactions and usage behaviors when using cosmetics are often not monitored in a timely and comprehensive manner, resulting in a lag in feedback information, which easily makes cosmetics manufacturers face greater challenges in product optimization and risk management. In order to improve the timeliness and accuracy of cosmetics safety monitoring, more and more companies and institutions have begun to focus on collecting user feedback information in real time through active monitoring and data analysis, and then dynamically evaluate the use effect and safety of cosmetics, so as to effectively monitor consumers' adverse skin reactions and behaviors when using cosmetics.

[0003] Existing technology, such as patent application publication number CN118427574A, discloses a method and system for analyzing and providing feedback on adverse reactions to cosmetics. This method relates to the field of cosmetics information management technology. The method includes collecting historical reports of adverse reactions to cosmetics from users, compiling statistics on key indicators of adverse reactions caused by user information and cosmetic information, preprocessing the historical report data, constructing a probability model for adverse reactions to cosmetics based on the key indicators, inputting the preprocessed historical report data into the probability model, simulating the probability model using Monte Carlo simulation, and then, combined with expert surveys, providing feedback to the user based on risk levels based on the risk simulation assessment results. This method addresses the difficulties associated with developing cosmetics risk management systems and the relative decrease in regulatory efficiency caused by the complex sources of cosmetics risk monitoring information, the scarcity of regulatory information analysis tools, and the inefficient storage and communication of information.

[0004] Based on the above solution, it is found that the limitations of the existing technology include at least the following problems: the existing technology does not fully take into account dynamic and real-time user feedback information, especially the different reactions that user groups of different age groups are prone to when using cosmetics, which can easily lead to insufficient accuracy and specificity in adverse reaction analysis, making it difficult to fully reflect the actual safety of cosmetics in different groups, and thus difficult to obtain accurate feedback and suggestions. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a method and system for analyzing and feedback on adverse reactions of cosmetics based on active monitoring. This solves the problem that the existing technology does not fully consider dynamic user feedback when analyzing adverse reactions of users of different age groups when using cosmetics, which easily leads to a lack of accuracy and pertinence in adverse reaction analysis.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a cosmetic adverse reaction analysis and feedback method based on active monitoring, comprising the following steps: obtaining usage monitoring feedback information of several age groups of set cosmetics, and inputting it into a pre-trained multimodal recognition model for predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; performing data analysis on the skin condition feedback dataset and the usage behavior evaluation feedback dataset for each age group of the set cosmetics, respectively, to obtain a skin adverse feedback index and a usage adverse feedback index for each age group of the set cosmetics; simultaneously obtaining an ingredient feedback safety index of the set cosmetics, and performing a comprehensive analysis based on the skin adverse feedback index and the usage adverse feedback index of each age group to obtain a comprehensive adverse feedback index for each age group of the set cosmetics; performing a judgment analysis on the comprehensive adverse feedback index of each age group of the set cosmetics and a preset comprehensive adverse feedback index threshold set, wherein the comprehensive adverse feedback index threshold set includes a first comprehensive adverse feedback index threshold, a second comprehensive adverse feedback index threshold, and a third comprehensive adverse feedback index threshold, and marking based on the judgment and analysis results, and generating corresponding feedback suggestions.

[0007] Furthermore, the multimodal recognition model is specifically a visual-text joint model, which includes a visual encoder layer, a text encoder layer, a joint embedding layer, and a regression output layer. The specific steps of obtaining a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of set cosmetics are as follows: in the visual encoder layer of the visual-text joint model, image feature extraction processing is performed on the usage monitoring feedback information of each age group of set cosmetics to obtain a high-dimensional image feature vector for each age group of set cosmetics; in the text encoder layer of the visual-text joint model, text feature extraction processing is performed on the usage monitoring feedback information of each age group of set cosmetics to obtain a text feature vector for each age group of set cosmetics; in the joint embedding layer of the visual-text joint model, The high-dimensional image feature vector and text feature vector of each age group of the set cosmetics are fused to obtain a joint feature vector of each age group of the set cosmetics; in the regression output layer of the visual-text joint model, the joint feature vector of each age group of the set cosmetics is subjected to regression prediction processing to obtain a skin state feedback dataset and a usage behavior evaluation feedback dataset of each age group of the set cosmetics; the skin state feedback dataset includes a skin oil secretion level value, an erythema index, a skin keratin thickness value, a skin surface moisture content value, a skin temperature gradient value, a melanin index, and a skin conductivity value; the usage behavior evaluation feedback dataset includes a usage frequency value, a usage dosage value, a duration value, a product mixing index, a comfort index, a stickiness index, a skin gloss index, and a stinging index.

[0008] Furthermore, the specific formula for calculating the comprehensive adverse feedback index of cosmetics for each age group is as follows:

[0009] ;

[0010] in, To set the cosmetic The comprehensive negative feedback index of each age group, 、 The first step to set cosmetics Skin adverse feedback index and usage adverse feedback index for each age group, To set the safety index of cosmetic ingredients feedback, 、 、 、 The following are the skin adjustment coefficient, usage adjustment coefficient, ingredient adjustment coefficient, and adverse interaction coefficient stored in the database. 1, 2, 3, ..., , The number of age groups.

[0011] Furthermore, the specific steps for obtaining the skin adverse feedback index of each age group of the set cosmetics are as follows: reading the skin oil secretion level value, erythema index, skin keratin thickness value, skin surface moisture content value, skin temperature gradient value, melanin index, and skin conductivity value in the skin state feedback data set for each age group of the set cosmetics; and obtaining the skin oil secretion level reference value, skin keratin thickness reference value, skin surface moisture content reference value, and environmental factors for each age group of the set cosmetics; comprehensively analyzing the skin oil secretion level reference value, skin keratin thickness reference value, skin surface moisture content reference value, skin oil secretion level value, erythema index, skin keratin thickness value, and skin surface moisture content value for each age group of the set cosmetics to obtain the skin damage index for each age group of the set cosmetics; and comprehensively analyzing the skin temperature gradient value, melanin index, skin conductivity value, and environmental factors for each age group of the set cosmetics to obtain the skin sensitivity index for each age group of the set cosmetics, and performing comprehensive analysis in combination with the skin damage index to obtain the skin adverse feedback index for each age group of the set cosmetics.

[0012] Furthermore, the specific formulas for calculating the skin damage index, skin sensitivity index, and skin adverse feedback index for each age group of the cosmetics are as follows:

[0013] ;

[0014] in, To set the cosmetic Skin damage index for each age group, 、 、 、 、 、 、 The first step to set cosmetics Skin oil secretion level value, skin oil secretion level reference value, erythema index, skin keratin thickness value, skin keratin thickness reference value, skin surface moisture content value, skin surface moisture content reference value, 、 、 、 The following are the oil secretion adjustment coefficient, erythema adjustment coefficient, skin keratin adjustment coefficient, and moisture content adjustment coefficient stored in the database. To set the cosmetic Skin sensitivity index for each age group, 、 、 、 The first step to set cosmetics Skin temperature gradient value, melanin index, skin conductivity value, environmental factors for each age group, 、 、 、 、 The following are the sensitive interaction coefficient, temperature gradient adjustment coefficient, melanin adjustment coefficient, skin conductivity value, conductivity adjustment coefficient, and environment adjustment coefficient stored in the database. To set the cosmetic Skin adverse feedback index for each age group, 、 、 、 The damage coefficient, damage adjustment coefficient, sensitivity coefficient, and sensitivity adjustment coefficient stored in the database are listed in order. , is a natural constant, 1, 2, 3, ..., , The number of age groups.

[0015] Furthermore, the specific steps for obtaining the adverse usage feedback index of the set cosmetics for each age group are as follows: reading the usage frequency value, usage dosage value, duration value, product mixing index, comfort index, stickiness index, skin gloss index, and stinging index in the usage behavior evaluation feedback data set of the set cosmetics for each age group; and comprehensively analyzing the usage frequency value, usage dosage value, duration value, and product mixing index of the set cosmetics for each age group to obtain the usage risk index of the set cosmetics for each age group; comprehensively analyzing the comfort index, stickiness index, skin gloss index, and stinging index of the set cosmetics for each age group to obtain the sensory evaluation index of the set cosmetics for each age group, and comprehensively analyzing in combination with the usage risk index to obtain the adverse usage feedback index of the set cosmetics for each age group.

[0016] Furthermore, the specific formula for calculating the sensory evaluation index and adverse feedback index of cosmetics for each age group is as follows:

[0017] ;

[0018] in, To set the cosmetic Sensory evaluation index of each age group, 、 、 、 The first step to set cosmetics Comfort index, skin gloss index, stickiness index, and tingling index for each age group. 、 、 They are the positive adjustment coefficient, negative adjustment coefficient, and difference adjustment coefficient stored in the database, To set the cosmetic The negative feedback index for each age group, To set the cosmetic The risk index for use in each age group, 、 、 、 The following are the behavior adjustment coefficient, sensory evaluation adjustment coefficient, behavior and sensory evaluation difference adjustment coefficient, and behavior and sensory evaluation interaction adjustment coefficient stored in the database. 1, 2, 3, ..., , The number of age groups.

[0019] Furthermore, the specific steps for obtaining the ingredient feedback safety index of the set cosmetics are as follows: obtaining the concentration value, concentration reference value, irritation score value, sensitization score value, and acne-causing score value of each ingredient of the set cosmetics; and conducting a comprehensive analysis of the concentration value, concentration reference value, irritation score value, sensitization score value, and acne-causing score value of each ingredient of the set cosmetics to obtain the ingredient feedback safety index of the set cosmetics.

[0020] Furthermore, the specific steps of generating corresponding feedback suggestions based on the judgment and analysis results are as follows: if the comprehensive adverse feedback index of the cosmetics for each age group is lower than or equal to the preset first comprehensive adverse feedback index threshold, it is marked as no adverse reaction, and a first feedback suggestion is generated; if the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset first comprehensive adverse feedback index threshold and lower than or equal to the preset second comprehensive adverse feedback index threshold, it is marked as a mild adverse reaction, and a second feedback suggestion is generated; if the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset second comprehensive adverse feedback index threshold and lower than or equal to the preset third comprehensive adverse feedback index threshold, it is marked as a moderate adverse reaction, and a third feedback suggestion is generated; if the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset third comprehensive adverse feedback index threshold, it is marked as a severe adverse reaction, and a fourth feedback suggestion is generated.

[0021] The system for analyzing and feedbacking adverse reactions to cosmetics based on active monitoring includes: a data acquisition module for acquiring usage monitoring feedback information of a set of cosmetics for several age groups, and inputting the information into a pre-trained multimodal recognition model for predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; a data analysis module for performing data analysis on the skin condition feedback dataset and the usage behavior evaluation feedback dataset for each age group of the set cosmetics, respectively, to obtain a skin adverse feedback index and a usage adverse feedback index for each age group of the set cosmetics; a comprehensive analysis module for simultaneously acquiring an ingredient feedback safety index of the set cosmetics, and performing comprehensive analysis in combination with the skin adverse feedback index and the usage adverse feedback index for each age group to obtain a comprehensive adverse feedback index for each age group of the set cosmetics; and a judgment and feedback module for performing judgment and analysis on the comprehensive adverse feedback index for each age group of the set cosmetics with a preset comprehensive adverse feedback index threshold set, wherein the comprehensive adverse feedback index threshold set includes a first comprehensive adverse feedback index threshold, a second comprehensive adverse feedback index threshold, and a third comprehensive adverse feedback index threshold, and marking based on the judgment and analysis results, and generating corresponding feedback suggestions.

[0022] The present invention has the following beneficial effects:

[0023] (1) This method of analyzing and feedback on adverse reactions to cosmetics based on active monitoring obtains usage monitoring feedback information of set cosmetics for each age group and inputs it into a pre-trained multimodal recognition model for predictive analysis, thereby updating the skin condition and usage evaluation behavior data of each age group in real time, thereby effectively capturing the immediate feedback that occurs when users of different age groups use cosmetics, and then providing targeted analysis, thereby accurately evaluating the adverse reactions of cosmetics and improving the safety and effectiveness of product use.

[0024] (2) This method of analyzing and feedback on adverse reactions of cosmetics based on active monitoring, through comprehensive analysis of the skin adverse feedback index, usage adverse feedback index and ingredient feedback safety index, obtains a comprehensive adverse feedback index for each age group, thereby enabling an in-depth understanding of the performance of cosmetics in different user groups and timely detection of existing adverse reactions, and then adjusts the product formula or provides personalized usage suggestions accordingly, thereby reducing the occurrence of adverse reactions and improving the market acceptance of the product.

[0025] (3) This method of analyzing and feedback on adverse reactions of cosmetics based on active monitoring can timely detect adverse reactions of cosmetics in specific groups by comparing the comprehensive adverse feedback index of each age group with the preset comprehensive adverse feedback index threshold, and generate personalized feedback suggestions, thereby effectively avoiding large-scale adverse reactions and providing real-time guidance for product optimization, thereby ensuring that consumers of different age groups have a safer and more effective use experience.

[0026] (4) The cosmetics adverse reaction analysis and feedback system based on active monitoring obtains usage monitoring feedback information from users of different age groups in real time and performs intelligent analysis through a pre-trained multimodal recognition model. This makes the usage feedback of cosmetics not limited to regular surveys or consumer complaints, but can obtain users' immediate responses at any time, thereby ensuring the accuracy and timeliness of feedback, and then quickly adjusting product formulas or optimizing usage suggestions, and ensuring that users reduce the occurrence of adverse reactions during use, thereby improving the market competitiveness of the product.

[0027] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of the cosmetic adverse reaction analysis and feedback method based on active monitoring of the present invention.

[0029] Figure 2 This is a flowchart of the specific steps for obtaining the adverse feedback index of the use of a set cosmetic for each age group in the cosmetic adverse reaction analysis and feedback method based on active monitoring of the present invention.

[0030] Figure 3 This is a block diagram of the cosmetic adverse reaction analysis and feedback system based on active monitoring of the present invention. DETAILED DESCRIPTION

[0031] See also Figure 1, an embodiment of the present invention provides a technical solution: a method for analyzing and feedback of adverse reactions of cosmetics based on active monitoring, comprising the following steps: obtaining usage monitoring feedback information of several (usage) age groups of set cosmetics (it should be noted here that the usage monitoring feedback text information is user feedback actively collected by relevant personnel or institutions, including feedback text information and feedback image information, and the data source includes but is not limited to text information, inspection information, image information, etc. actively reported or monitored by consumers or medical professionals in real time when using cosmetics), and inputting it into a pre-trained multimodal recognition model for predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; performing data analysis on the skin condition feedback dataset and the usage behavior evaluation feedback dataset for each age group of the set cosmetics respectively to obtain skin adverse reaction datasets for each age group of the set cosmetics. Feedback index, usage adverse feedback index; at the same time, obtain the ingredient feedback safety index of the set cosmetics, and conduct a comprehensive analysis in combination with the skin adverse feedback index and usage adverse feedback index of each age group to obtain the comprehensive adverse feedback index of each age group of the set cosmetics; judge and analyze the comprehensive adverse feedback index of each age group of the set cosmetics with the preset comprehensive adverse feedback index threshold set, the comprehensive adverse feedback index threshold set includes the first comprehensive adverse feedback index threshold (the lowest adverse feedback threshold, used to identify whether there is an adverse reaction during the use of cosmetics), the second comprehensive adverse feedback index threshold (the medium adverse feedback threshold, used to identify moderate adverse reactions during the use of cosmetics), and the third comprehensive adverse feedback index threshold (the highest adverse feedback threshold, used to judge whether the cosmetics have serious adverse reactions), and mark them based on the judgment and analysis results, and generate corresponding feedback suggestions.

[0032] The specific formula for calculating the comprehensive adverse feedback index for each age group of a given cosmetic is as follows:

[0033] ;

[0034] in, To set the cosmetic The comprehensive negative feedback index of each age group, To set the cosmetic Skin adverse feedback index for each age group, is the skin adjustment coefficient stored in the database, To set the cosmetic The negative feedback index for each age group, is the usage adjustment factor stored in the database, To set the safety index of cosmetic ingredients feedback, is the component adjustment coefficient stored in the database, is the bad interaction coefficient stored in the database, 1, 2, 3, ..., , The number of age groups.

[0035] It needs to be explained that the formula This item is used to adjust the superimposed effect of skin adverse feedback index, usage adverse feedback index, and ingredient feedback safety index to avoid the comprehensive adverse feedback index being too high or too low.

[0036] 、 、 、 It can be obtained through the following steps: using historical data, combined with the skin adverse feedback index, usage adverse feedback index, and ingredient feedback safety index, to conduct statistical regression analysis, quantify the specific impact of each factor on the comprehensive adverse feedback index, and thus fit the initial weight value; secondly, using the sensitivity analysis method, adjust the value range of each coefficient, observe its impact on the comprehensive adverse feedback index, ensure the stability and rationality of the model, and based on the set cosmetic characteristics and actual conditions, correct and optimize the preliminary fitting coefficients, and finally determine the coefficient value applicable to the specific cosmetics.

[0037] Specifically, the multimodal recognition model is a visual-text joint model (i.e., Vision-Language Models). The visual-text joint model includes a visual encoder layer, a text encoder layer, a joint embedding layer, and a regression output layer. The specific steps of obtaining the skin condition feedback dataset and the usage behavior evaluation feedback dataset for each age group of the set cosmetics are as follows: in the visual encoder layer of the visual-text joint model, the usage monitoring feedback information of each age group of the set cosmetics (i.e., the feedback image information in the usage monitoring feedback information) is subjected to image feature extraction processing (i.e., low-level to high-level features, such as edges, textures, and shapes, are extracted through convolution and pooling operations of the convolution layer, and activation functions such as ReLU are used to increase nonlinear capabilities. Then, methods such as maximum pooling are used to reduce the dimension of the feature map, reduce computational complexity, and retain significant features. The features after convolution and pooling are then flattened and mapped to a high-dimensional feature space through a fully connected layer to learn the overall information. Finally, a fixed-dimensional feature vector is output through the fully connected layer to represent the high-dimensional features of the image, such as skin color, texture, erythema, etc.), and a high-dimensional image feature vector for each age group of the set cosmetics is obtained; in the text encoder layer of the visual-text joint model, the set The usage monitoring feedback information for each age group of cosmetics (i.e., the feedback text information in the usage monitoring feedback information) is subjected to text feature extraction processing (i.e., each word or subword in the text is converted into a high-dimensional vector, such as using Word2Vec or GloVe. Based on Transformer: the self-attention mechanism is used to learn the lexical relationships in the text, and a global representation of the text is obtained by averaging, weighting, or concatenating word vectors. The vector corresponding to the CLS tag output by a model such as BERT is used to represent the text features. Finally, a pooling operation (maximum pooling / average pooling, etc.) is used to extract a fixed-length vector representation). The text feature vector for each age group of the set cosmetics is obtained. In the joint embedding layer of the joint visual-text model, the high-dimensional image feature vector and text feature vector for each age group of the set cosmetics are fused (i.e., the image feature vector and text feature vector are concatenated in a certain dimension to form a larger vector, and the image and text features are weighted summed to obtain a comprehensive feature vector. The self-attention mechanism is used to combine the image and text features, the importance of each feature is calculated, and the weighted fusion is performed to obtain the final joint feature vector). The joint feature vector for each age group of the set cosmetics is obtained.In the regression output layer of the visual text joint model, the joint feature vector of each age group of the set cosmetics is subjected to regression prediction processing (that is, the joint feature vector is input into multiple fully connected layers, and multiple fully connected layers are used. Each task corresponds to an output branch. Skin physiological state: such as erythema, skin oil secretion level, etc., output continuous values, usage behavior: such as frequency of use, duration, etc., output continuous values, establish an independent output layer for each task, each output layer consists of several neurons, and the output dimension is determined according to the prediction indicator, such as skin physiological state: output values related to skin state, For example, the skin oil secretion level value, the joint feature vector will contain the skin surface oil features from the image, such as glossiness, oiliness, etc., as well as user feedback in the text, such as "greasy face", "face feels dry", etc. The image features will extract the oil secretion situation on the skin surface, such as the oiliness and glossiness on the skin, etc. The text features will extract emotions and descriptions, such as "greasy" and "dry", which reflect the oiliness of the skin. These image and text features are combined and processed by the regression of the fully connected layer to output a continuous value, indicating the oil secretion level of the skin, and output: Skin oil secretion level value , is a range value, such as 0 to 1, indicating the oiliness of the skin; Usage evaluation behavior: outputs values related to usage behavior, such as comfort index. Image features are related to the reaction of the skin surface, such as oiliness, redness, etc. The text describes the user's feelings when using it, such as "feeling light" and "no burden after use". Image features assist in prediction through the surface reaction of the skin, such as dryness and greasiness, and text features are weighted to evaluate comfort through sentiment analysis, such as "no irritation" and "very comfortable". Using this information through regression output, a continuous comfort score is obtained and output The comfort index (a range of values, such as 0 to 1, indicating the user's comfort when using cosmetics) is used to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the specified cosmetics. The skin condition feedback dataset includes skin oil secretion level, erythema index, skin keratin thickness, skin surface moisture content, skin temperature gradient, melanin index, and skin conductivity. The usage behavior evaluation feedback dataset includes usage frequency, dosage, duration, product mixing index, comfort index, stickiness index, skin gloss index, and stinging index.

[0038] Among them, the skin oil secretion level value is the amount of sebum secreted on the skin surface, which is used to reflect the oiliness of the skin. Too high or too low oil secretion levels can easily lead to skin barrier damage.

[0039] The erythema index is a measure of the degree of erythema that appears on the skin surface. It is used to reflect the damage or allergic reaction of the skin after being stimulated by external stimuli. A higher erythema index indicates that the skin has a stronger irritation reaction to cosmetics and there is a certain risk of skin damage.

[0040] The stratum corneum thickness value is the thickness of the protective barrier layer on the surface of the skin. It is used to reflect the barrier function of the skin. A stratum corneum that is too thin or too thick will affect the health of the skin.

[0041] The skin surface moisture content value is the moisture content on the skin surface, which is used to reflect the skin's hydration and moisturizing ability. Low moisture content often indicates that the skin lacks effective protection and nourishment and is susceptible to damage. Excessive moisture content in the skin can easily lead to insufficient oil secretion of the skin, thereby causing damage to the barrier function. Once the barrier is damaged, the skin is susceptible to invasion by harmful external substances.

[0042] The skin temperature gradient value is the distribution of skin surface temperature. It is used to reflect that when the temperature gradient is small, the skin surface is relatively stable, the blood circulation is good, the skin may be more tolerant to external stimuli, and have fewer sensitive reactions.

[0043] The melanin index is a measure of the distribution of melanin in the skin. It reflects the degree of skin pigmentation, affects the skin's color and response to ultraviolet rays. People with a low melanin index have more sensitive skin and are easily damaged by ultraviolet rays.

[0044] The skin conductivity value is the skin's ability to conduct electric current. It is used to reflect that when the conductivity is high, the skin surface is moist and easily has better hydration. Good hydration helps maintain the skin's barrier function, thereby reducing skin sensitivity.

[0045] The frequency of use value is the frequency of daily use of cosmetics. It is used to reflect that frequent use will increase the long-term impact of the product on the skin, such as cumulative effects, which may easily lead to allergies or other adverse skin reactions.

[0046] The use dose value is the amount of product applied per use and is used to reflect that higher doses of cosmetics are likely to cause greater skin risks.

[0047] The duration value is the duration from each use of the product to makeup removal. It is used to reflect that the product ingredients will act more deeply into the skin, affecting the skin barrier and function.

[0048] The product mixing index is the degree to which multiple cosmetics are used in daily skin care. It refers to the user's use of multiple skin care products at the same time. It is used to reflect that when mixing multiple products, it is easy to cause interactions between ingredients and increase the risk of skin discomfort or allergies.

[0049] The comfort index is the user's sense of comfort when using cosmetics. A low comfort index indicates discomfort, excessive oiliness, stickiness, etc. during use, which directly affects the natural feeling of the skin and easily leads to allergic or uncomfortable feedback.

[0050] The stickiness index represents the degree of stickiness on the skin after a user applies cosmetics.

[0051] The skin glossiness index measures the intensity of the glossiness on the skin surface.

[0052] The stinging index represents the stinging sensation when a user applies cosmetics.

[0053] Moreover, the parameters in the skin condition feedback dataset and the usage behavior evaluation feedback dataset are unitless parameters, and their value ranges are both between 0 and 1.

[0054] The visual encoder layer includes: a convolutional layer, an activation function, a pooling layer, a fully connected layer, and an output layer.

[0055] The convolutional layer is used to extract local features of an image, such as edges, textures, shapes, etc.

[0056] The activation function is used to introduce non-linearity so that the model can process more complex features. The commonly used ReLU (Rectified Linear Unit) activation function increases non-linearity by setting negative values to 0.

[0057] The pooling layer is used to reduce the spatial dimension of the feature map, reduce computational complexity, and retain important features at the same time.

[0058] The fully connected layer is used to map the local features extracted by the convolutional layer and the pooling layer to a global feature space.

[0059] The output layer is used to output a high-dimensional feature vector at the end of the visual encoder layer as the final representation of the image.

[0060] The text encoder layer includes: a text preprocessing layer, a word embedding layer, a recurrent network layer, a Transformer layer, a self-attention layer, a text pooling layer, a text fully connected layer, and a text output layer.

[0061] The text preprocessing layer is used to clean and standardize the input text to prepare data for subsequent feature extraction. The specific steps are: tokenization, splitting the text into words or subwords and removing stop words, such as deleting common words with no practical meaning (such as "de", "shi", etc.), then lowercase conversion, converting all letters to lowercase to reduce unnecessary repetition, and finally stemming or lemmatization to restore the words to their basic forms.

[0062] The word embedding layer is used to map each word or subword to a high-dimensional vector so that it can better represent the semantic information of the word.

[0063] The recurrent network layer is used to capture the temporal and dependency relationships in the text, and is especially suitable for processing long texts.

[0064] The Transformer layer is used to capture the relationship between words based on the self-attention mechanism, which is particularly suitable for capturing long-distance dependencies in text.

[0065] The self-attention layer is used to calculate the relationship between words in a sentence, helping the model better understand the dependencies between words in the text.

[0066] The text pooling layer is used to extract fixed-length vector representations from word vector sequences to reduce the computational complexity of the model.

[0067] The text fully connected layer is used to map the features extracted from the text to the final feature space.

[0068] The text output layer is used to finally generate a vector representing the text, which is usually used for subsequent tasks such as sentiment analysis, text classification, or regression tasks.

[0069] The joint embedding layer is used to fuse the feature vectors from the visual encoder and the text encoder to form a joint feature representation that can contain both image and text information.

[0070] The regression output layer includes a regression fully connected layer, a linear activation function, and output layer neurons.

[0071] The regression fully connected layer is used to map the joint feature vector or the features processed by the previous layer to the target space of the regression task, that is, to linearly combine the joint feature vector with the weight matrix, and then perform a nonlinear transformation through the activation function to output the feature vector or numerical value required for the regression task.

[0072] The linear activation function is used to transform the output of the network from the feature space to the numerical value required for the regression task.

[0073] Output layer neurons are used to map the number of neurons in the output layer to the target dimension of the regression task. For example, predicting each value of a skin physiological state (such as erythema, allergic reaction, etc.) may require a separate output neuron. When there are multiple regression tasks (such as predicting skin physiological state and usage behavior simultaneously), each task is assigned a neuron.

[0074] The specific process of pre-training is as follows:

[0075] Obtain several sets of feedback information (including images and corresponding text descriptions, such as image annotations, labels, user comments, etc.) and divide them into training sets and validation sets.

[0076] Training based on the training set:

[0077] Number of training cycles: Set the number of training cycles (for example, 50 or 100 training cycles).

[0078] Forward propagation: Input the training set, extract features through the visual encoder and text encoder respectively, then fuse the features through the joint embedding layer, and finally input them into the regression layer for prediction.

[0079] Calculate loss: Calculate the loss function (such as mean squared error loss, contrastive loss, etc.) based on the model output and the true label.

[0080] Backpropagation: Calculate the gradient through the backpropagation algorithm and update the model parameters. Use optimization algorithms such as Adam optimizer or SGD to update the weights.

[0081] Optimizer selection: Commonly used optimizers include Adam and RMSProp, which can effectively optimize multi-task regression models.

[0082] Learning rate adjustment: Use learning rate decay or scheduling strategies to avoid overfitting or non-convergence caused by too fast training.

[0083] Batch size adjustment: Choose an appropriate batch size based on hardware limitations and the size of the dataset. Generally, a larger batch size can speed up training but may cause insufficient memory.

[0084] Validation set: Regular evaluation is performed on the validation set to monitor the model's performance in predicting skin physiological status and usage behavior.

[0085] Overfitting monitoring: Monitor changes in training error and validation error to prevent the model from overfitting. An early stopping strategy can be used to stop training when the validation error stops decreasing.

[0086] Regression task evaluation: Use indicators such as R2 value, mean square error (MSE), and mean absolute error (MAE) to evaluate the model's prediction performance for skin physiological status and usage behavior.

[0087] Test set evaluation: Use the test set to perform a final evaluation of the model to verify the generalization ability of the model.

[0088] Task-specific fine-tuning: Fine-tuning for specific tasks (such as skin physiological state prediction and usage behavior prediction) is performed based on different application scenarios or task requirements.

[0089] Transfer learning: If there is a pre-trained model in a related field, you can use transfer learning to fine-tune the model on the target task dataset.

[0090] In this embodiment, through the joint modeling of visual and textual information, the effects of cosmetics on the skin and user feedback can be more comprehensively captured. Visual information provides specific manifestations of the skin (such as oil secretion, erythema, etc.), while textual information provides users' subjective feelings and descriptions (such as "greasy", "comfortable", etc.). The fusion of these two types of information helps the model to more accurately understand the skin condition and usage behavior, thereby improving the accuracy of adverse reaction analysis. Secondly, through multi-task learning, the model can not only predict the physiological state of the skin (such as oil secretion, erythema, etc.), but also evaluate the user's usage behavior (such as comfort, frequency, etc.), thereby optimizing multiple objectives at the same time, thereby improving the overall efficiency and applicability of the model. At the same time, through automated analysis of computer vision and natural language processing, human intervention can be reduced, thereby improving the efficiency of the entire evaluation process, allowing manufacturers to obtain feedback on skin condition and usage behavior in a short period of time, so as to make quick adjustments and improve the use effect of cosmetics and user experience.

[0091] Specifically, the specific steps for obtaining the skin adverse feedback index for each age group of the set cosmetics are as follows: reading the skin oil secretion level value, erythema index, skin keratin thickness value, skin surface moisture content value, skin temperature gradient value, melanin index, and skin conductivity value in the skin state feedback data set for each age group of the set cosmetics; and obtaining the skin oil secretion level reference value, skin keratin thickness reference value, skin surface moisture content reference value, and environmental factors for each age group of the set cosmetics; comprehensively analyzing the skin oil secretion level reference value, skin keratin thickness reference value, skin surface moisture content reference value, skin oil secretion level value, erythema index, skin keratin thickness value, and skin surface moisture content value for each age group of the set cosmetics to obtain the skin damage index for each age group of the set cosmetics; and comprehensively analyzing the skin temperature gradient value, melanin index, skin conductivity value, and environmental factors for each age group of the set cosmetics to obtain the skin sensitivity index for each age group of the set cosmetics, and performing comprehensive analysis in combination with the skin damage index to obtain the skin adverse feedback index for each age group of the set cosmetics.

[0092] Among them, the reference value of skin oil secretion level, the reference value of skin keratin thickness, and the reference value of skin surface moisture content can all be obtained through the skin database, and then standardized respectively.

[0093] The specific steps for obtaining environmental factors are as follows:

[0094] The humidity value, air quality value and corresponding humidity reference value and air quality reference value of each population in each age group are obtained, and a ratio analysis is performed. A weighted processing is performed based on the ratio analysis results, and a mean analysis is performed based on the weighted processing results to obtain the environmental factors for each age group.

[0095] The humidity value and the air quality value can be obtained from the weather station monitoring report stored in the database, and the humidity reference value and the air quality reference value can be obtained from the standard reference data released by the meteorological bureau stored in the database.

[0096] The specific formulas for calculating the skin damage index, skin sensitivity index, and skin adverse feedback index for each age group of cosmetics are as follows:

[0097] ;

[0098] in, To set the cosmetic Skin damage index for each age group, To set the cosmetic Skin oil secretion levels for each age group, To set the cosmetic Reference values for skin oil secretion levels for each age group: is the oil secretion adjustment coefficient stored in the database, To set the cosmetic Erythema index of each age group, is the erythema adjustment coefficient stored in the database, To set the cosmetic Skin keratin thickness values for each age group, To set the cosmetic Reference values for skin keratin thickness for each age group: is the skin keratin adjustment coefficient stored in the database, To set the cosmetic Skin surface moisture content values for each age group, To set the cosmetic Reference values for skin surface moisture content for different age groups: is the moisture content adjustment coefficient stored in the database, To set the cosmetic Skin sensitivity index for each age group, is the sensitive interaction coefficient stored in the database, To set the cosmetic Skin temperature gradient values for each age group, is the temperature gradient adjustment coefficient stored in the database, To set the cosmetic Melanin index of each age group, is the melanin adjustment coefficient stored in the database, To set the cosmetic Skin conductivity values for each age group, is the conductivity adjustment coefficient stored in the database, To set the cosmetic Environmental factors of different age groups, is the environmental adjustment coefficient stored in the database, To set the cosmetic Skin adverse feedback index for each age group, is the damage coefficient stored in the database, is the damage adjustment factor stored in the database, is the sensitivity coefficient stored in the database, is the sensitivity adjustment coefficient stored in the database, , is a natural constant and in this embodiment takes a value of 2.71. 1, 2, 3, ..., , The number of age groups.

[0099] What needs to be explained is that 、 、 、 It can be obtained through the following steps: using historical data, combined with skin oil secretion level values, erythema index, skin keratin thickness values, and skin surface moisture content values, statistical regression analysis is performed to quantify the specific impact of each factor on the skin damage index, thereby fitting the initial weight value; secondly, using the sensitivity analysis method, adjust the value range of each coefficient, observe its impact on the skin damage index, ensure the stability and rationality of the model, and based on the set cosmetic characteristics and actual conditions, correct and optimize the preliminary fitting coefficients, and finally determine the coefficient value applicable to the specific cosmetics.

[0100] 、 、 、 、 It can be obtained through the following steps: Based on historical data, determine the initial impact weight of each variable (such as skin temperature gradient value, melanin index, etc.) on the skin sensitivity index through statistical regression analysis. Then, use the sensitivity analysis method to adjust the value range of the coefficient to evaluate the stability and applicability of these parameters to the formula output. Next, further fit the weights through model optimization (such as multi-objective optimization) to ensure that the formula can accurately reflect the skin sensitivity state, and fine-tune the coefficients based on the characteristics of different cosmetics to ensure that it is suitable for specific skin sensitivity assessment needs.

[0101] 、 It can be obtained through the following steps: read the skin sensitivity index and skin damage index of each age group of the set cosmetics, and perform mean processing to obtain the mean skin sensitivity index and the mean skin damage index of the set cosmetics, perform sum analysis to obtain the bad skin sum value, and perform proportion analysis on the mean skin sensitivity index and the mean skin damage index of the set cosmetics and the bad skin sum value respectively, and use the proportion results as the corresponding coefficients.

[0102] 、 It can be obtained through the following steps: using historical data, combined with the skin damage index and skin sensitivity index, to conduct statistical regression analysis, quantify the specific impact of each factor on the skin adverse feedback index, and thus fit the initial weight value; secondly, using the sensitivity analysis method, adjust the value range of each coefficient, observe its impact on the skin adverse feedback index, ensure the stability and rationality of the model, and based on the set cosmetic characteristics and actual conditions, correct and optimize the preliminary fitting coefficients, and finally determine the coefficient value applicable to the specific cosmetics.

[0103] The specific implementation example of calculating the skin adverse feedback index for the first age group (e.g., 18-22 years old) of a cosmetic is as follows. The existing data is as follows:

[0104] The skin oil secretion level value for the first age group of cosmetics is set to: 0.86.

[0105] The erythema index of the first age group of cosmetics is set to: 0.32.

[0106] Set the skin keratin thickness value for the first age group of cosmetics to: 0.38.

[0107] The skin surface moisture content value for the first age group of cosmetics is set to: 0.73.

[0108] Set the skin temperature gradient value of the first age group of cosmetics to: 0.53.

[0109] The melanin index for the first age group of cosmetics is set to: 0.34.

[0110] The skin conductivity value of the first age group of cosmetics is set to: 0.59.

[0111] The environmental factor for the first age group of cosmetics is approximately: 0.64.

[0112] The reference value for skin oil secretion level for the first age group of cosmetics is set as: 0.75.

[0113] The reference value for skin keratin thickness for the first age group of cosmetics is set as: 0.50.

[0114] The reference value for skin surface moisture content in the first age group of cosmetics is set as: 0.65.

[0115] The oil secretion adjustment coefficient stored in the database is approximately: 0.57.

[0116] The erythema adjustment factor stored in the database is approximately: 0.18.

[0117] The skin keratin adjustment coefficient stored in the database is approximately: 0.61.

[0118] The moisture content adjustment factor stored in the database is approximately: 0.52.

[0119] The sensitive interaction coefficient stored in the database is approximately: 1.25.

[0120] The temperature gradient adjustment coefficient stored in the database is approximately: 0.38.

[0121] The melanin adjustment factor stored in the database is approximately: 0.51.

[0122] The conductivity adjustment factor stored in the database is approximately: 0.74.

[0123] The environmental adjustment factor stored in the database is approximately: 0.47.

[0124] The damage factor stored in the database is approximately: 0.48.

[0125] The damage adjustment factor stored in the database is approximately: 0.26.

[0126] The sensitivity coefficient stored in the database is approximately: 0.52.

[0127] The sensitivity adjustment coefficient stored in the database is approximately: 0.21.

[0128] Substituting the above data into the specific formula for calculating the skin damage index, skin sensitivity index, and skin adverse feedback index for each age group of the cosmetics, we obtain:

[0129] The skin damage index of the first age group of cosmetics is set to ≈0.58.

[0130] The skin sensitivity index of the first age group for cosmetics is set to ≈0.61.

[0131] The skin adverse feedback index of cosmetics for the first age group is set to ≈0.50.

[0132] In this embodiment, by comprehensively considering multiple skin physiological indicators such as the skin's oil secretion level, erythema index, stratum corneum thickness, moisture content, temperature gradient, etc., it is possible to more comprehensively reflect the overall impact of cosmetics on the skin, rather than relying solely on a single indicator, thereby helping to accurately capture the advantages and disadvantages of the product, and thus avoid erroneous evaluations caused by local data imbalance. Secondly, feedback data from different age groups are taken into account, so as to conduct personalized analysis based on the skin characteristics of different age groups, and by subdividing the skin reactions of different age groups, it is possible to optimize the recommendation and use effects of cosmetics. At the same time, incorporating environmental factors into the analysis can better reflect the adaptability of products under different climate, temperature and humidity conditions. For example, a dry environment may aggravate the loss of water in the skin, or the problem of oil secretion in the skin in a humid and hot environment, thereby helping to provide more realistic skin reaction feedback. Finally, through scientific comprehensive analysis methods, adverse skin reactions can be quantified, and traditional perceptual evaluations can be converted into data-based feedback, which is convenient for improving the market adaptability of products.

[0133] Specifically, if Figure 2 As shown, the specific steps for obtaining the adverse feedback index of the set cosmetics for each age group are as follows: reading the usage frequency value, usage dosage value, duration value, product mixing index, comfort index, stickiness index, skin gloss index, and stinging index in the usage behavior evaluation feedback data set of the set cosmetics for each age group; and performing a comprehensive analysis (i.e., weighted processing) on the usage frequency value, usage dosage value, duration value, and product mixing index of the set cosmetics for each age group to obtain the usage risk index of the set cosmetics for each age group; performing a comprehensive analysis on the comfort index, stickiness index, skin gloss index, and stinging index of the set cosmetics for each age group to obtain the sensory evaluation index of the set cosmetics for each age group, and performing a comprehensive analysis in combination with the usage risk index to obtain the adverse feedback index of the set cosmetics for each age group.

[0134] The specific formula for calculating the sensory evaluation index and adverse feedback index of cosmetics for each age group is as follows:

[0135] ;

[0136] in, To set the cosmetic Sensory evaluation index of each age group, To set the cosmetic The comfort index of each age group, To set the cosmetic Skin glossiness index for each age group, is the positive adjustment coefficient stored in the database, To set the cosmetic The stickiness index of each age group, To set the cosmetic The pain index of each age group, is the negative adjustment coefficient stored in the database, is the difference adjustment coefficient stored in the database, To set the cosmetic The negative feedback index for each age group, To set the cosmetic The risk index for use in each age group, is the behavior adjustment coefficient stored in the database, is the sensory evaluation adjustment coefficient stored in the database, is the difference adjustment coefficient between behavior and sensory evaluation stored in the database, is the interaction adjustment coefficient between behavior and sensory evaluation stored in the database, 1, 2, 3, ..., , The number of age groups.

[0137] It needs to be explained that the formula This item is used to control the impact of the difference between behavior and sensory evaluation and the interaction effect between behavior and sensory evaluation on the use of adverse feedback index.

[0138] 、 、 It can be obtained through the following steps: Based on historical data, determine the initial impact weight of each variable (comfort index, stickiness index, skin gloss index, and tingling index) on the sensory evaluation index through statistical regression analysis. Then, use the sensitivity analysis method to adjust the value range of the coefficient to evaluate the stability and applicability of these parameters to the formula output. Next, further fit the weights through model optimization (such as machine learning algorithms) to ensure that the formula can accurately reflect the sensory evaluation.

[0139] 、 、 、 It can be obtained through the following steps: using historical data, combined with the risk index and sensory evaluation index, to conduct statistical regression analysis, quantify the specific impact of each factor on the use of adverse feedback index, and thus fit the initial weight value; secondly, using the sensitivity analysis method, adjust the value range of each coefficient, observe its impact on the use of adverse feedback evaluation results, and ensure the stability and rationality of the model.

[0140] In this implementation, a comprehensive analysis of factors such as frequency of use, dosage, duration, and the misuse index can help identify potential risks of cosmetics under different usage conditions, thereby providing important information on product safety and applicability. This, in particular, helps prevent adverse reactions that may result from excessive or improper use for consumers of different age groups. Furthermore, the sensory experiences of different age groups (such as comfort, stickiness, and stinging) are considered, and feedback from consumers of different age groups is meticulously segmented. This helps manufacturers adjust formulations or usage instructions based on the preferences of the target population, thereby improving the market adaptability of products. Furthermore, by incorporating adjustment terms for "behavior-sensory evaluation differences" and "behavior-sensory evaluation interaction effects" into the formula, it helps to more accurately quantify the impact of different user behaviors on sensory experience, thereby precisely controlling this impact, thereby avoiding errors and improving the reliability of the assessment. Finally, through regression analysis and sensitivity analysis based on historical data, the impact of various indicators on the adverse user feedback index can be scientifically evaluated, enabling cosmetics manufacturers to adjust product strategies based on specific feedback.

[0141] Specifically, the specific steps for obtaining the ingredient feedback safety index of the set cosmetics are as follows: obtain the concentration value, concentration reference value, irritation score value, sensitization score value, and acne-causing score value of each ingredient of the set cosmetics; and conduct a comprehensive analysis of the concentration value, concentration reference value, irritation score value, sensitization score value, and acne-causing score value of each ingredient of the set cosmetics to obtain the ingredient feedback safety index of the set cosmetics.

[0142] The specific formula for calculating the ingredient feedback safety index of cosmetics is as follows:

[0143] ;

[0144] in, To set the safety index of cosmetic ingredients feedback, To set the cosmetic The concentration of the component, To set the cosmetic Reference concentration values of the components, is the concentration adjustment coefficient stored in the database, To set the cosmetic The irritation rating of the ingredients, is the stimulation gain coefficient stored in the database, is the stimulus adjustment coefficient stored in the database, To set the cosmetic The allergenicity score of each ingredient, is the sensitization gain coefficient stored in the database, is the sensitization adjustment coefficient stored in the database, To set the cosmetic The comedogenicity score of each ingredient, is the acne gain coefficient stored in the database, is the acne adjustment coefficient stored in the database, 1, 2, 3, ..., , is the number of components.

[0145] What needs to be explained is that 、 、 、 、 、 、 It can be obtained through the following steps: Based on the historical monitoring data of the region, determine the initial impact weight of each variable (concentration value, irritation score value, allergenicity score value, acnegenicity score value) on the ingredient feedback safety index through statistical regression analysis, then use the sensitivity analysis method to adjust the value range of the coefficient to evaluate the stability and applicability of these parameters to the formula output, and then further fit the weights through model optimization (such as machine learning algorithm or multi-objective optimization) to ensure that the formula can accurately reflect the health status of the actual ingredient feedback.

[0146] The concentration value is obtained from the formula book of the cosmetic stored in the database.

[0147] Concentration reference values are obtained through industry standards and regulations stored in the database.

[0148] Irritation score values, sensitization score values, and comedogenicity score values can all be obtained through toxicology databases (e.g., PubChem).

[0149] In this embodiment, by quantifying the concentration value, irritation, sensitization, acnegenicity and other factors of each ingredient and combining them with historical monitoring data, a comprehensive score of each ingredient on product safety can be provided, thereby scientifically evaluating the adverse reactions caused by the potential hazards of cosmetic ingredients. By comprehensively considering multiple factors such as ingredient concentration, irritation, sensitization, etc., manufacturers can more finely adjust and optimize the ingredient ratio of the product. For sensitive skin or special populations, safer and milder formulas can be formulated, thereby improving the product's scope of application and consumer satisfaction. Secondly, through regression analysis and sensitivity analysis, the impact of each parameter on the ingredient safety assessment can be quantified, and adjustments can be made based on these impact weights, thereby more accurately identifying ingredients that may have a negative impact on the skin and providing specific improvement directions for the R&D team. Finally, through comprehensive safety evaluation and optimization, it can ensure that the product is harmless to the consumer's skin, reduce the occurrence of adverse reactions such as allergies and irritation, thereby enhancing consumer trust and improving the product's market reputation.

[0150] Specifically, the specific steps for generating corresponding feedback suggestions based on the judgment and analysis results are as follows: if the comprehensive adverse feedback index of the cosmetics for each age group is set to be lower than or equal to a preset first comprehensive adverse feedback index threshold, then (this age group) is marked as having no adverse reaction, and a first feedback suggestion is generated (i.e., providing the user with feedback suggestions such as continuing to use the cosmetics, suggesting that the user continue to use the cosmetics because no adverse reaction has occurred during use in this age group; maintaining a normal frequency of use, according to the recommended usage method of the product, maintaining a normal frequency of use; encouraging feedback, encouraging the user to continue to provide feedback on the experience during continued use, and ensuring the long-term safety of the cosmetics on the skin); if the comprehensive adverse feedback index of the cosmetics for each age group is set to be higher than the preset first comprehensive adverse feedback index threshold and lower than or equal to the preset second comprehensive adverse feedback index threshold, then (this age group) is marked as having a mild adverse reaction, and a second feedback suggestion is generated (i.e., providing the user with feedback suggestions such as adjusting the usage method, suggesting reducing the frequency of use or reducing the amount of use, or appropriately changing the time and method of use; paying attention to skin conditions, suggesting that the user pay close attention to skin reactions, and immediately stop using the product if further discomfort occurs, and consulting a doctor). Consult a professional doctor for advice); if the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset second comprehensive adverse feedback index threshold and lower than or equal to the preset third comprehensive adverse feedback index threshold, then (that age group) is marked as a moderate adverse reaction, and a third feedback suggestion is generated (i.e., an in-depth review of ingredients and formulas is provided to the manufacturer. All ingredients of the cosmetics should be reviewed, especially those that may cause moderate adverse reactions, to assess whether the concentration of these ingredients needs to be reduced or replaced with milder ingredients); if the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset third comprehensive adverse feedback index threshold, then (that age group) is marked as a severe adverse reaction, and a fourth feedback suggestion is generated (i.e., recommendations are provided to the manufacturer to stop sales and withdraw the cosmetics from the market, to stop sales and consider withdrawing the batch of cosmetics from the market to prevent more users from being affected. At the same time, it is recommended that the manufacturer conduct an ingredient and formula review to identify specific ingredients that may cause severe adverse reactions, especially negative effects on sensitive skin or specific groups; improve product research and development and testing processes, strengthen the product research and development process, and increase multi-stage safety testing).

[0151] In this implementation plan, personalized feedback suggestions are formulated based on the comprehensive adverse feedback index of different age groups to help users adjust the usage of cosmetics according to their own circumstances, maximize the avoidance of adverse reactions, ensure skin safety, and for different degrees of adverse reactions, users can quickly obtain guidance to avoid further discomfort, take timely measures to alleviate symptoms, and enhance user confidence and experience. Secondly, by analyzing the feedback data of each age group and conducting judgment analysis with preset thresholds, manufacturers can promptly identify possible risks in the product, especially the reactions of sensitive groups, and promote improvements in product design and research and development. Finally, through a scientific and phased feedback system, not only can user safety and continuous product optimization be ensured, but also the manufacturer's sense of responsibility and product quality control can be enhanced, ultimately forming a virtuous product improvement and market feedback cycle.

[0152] See also Figure 3 The embodiment of the present invention provides a technical solution: a cosmetic adverse reaction analysis and feedback system based on active monitoring, comprising: a data acquisition module for acquiring usage monitoring feedback information of a set of cosmetics for several age groups, and inputting the information into a pre-trained multimodal recognition model for predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; a data analysis module for performing data analysis on the skin condition feedback dataset and the usage behavior evaluation feedback dataset for each age group of the set cosmetics, respectively, to obtain a skin adverse feedback index and a usage adverse feedback index for each age group of the set cosmetics; a comprehensive analysis module for simultaneously acquiring an ingredient feedback safety index of the set cosmetics, and performing comprehensive analysis based on the skin adverse feedback index and the usage adverse feedback index for each age group to obtain a comprehensive adverse feedback index for each age group of the set cosmetics; and a judgment and feedback module for performing judgment and analysis on the comprehensive adverse feedback index of each age group of the set cosmetics against a preset comprehensive adverse feedback index threshold set, the comprehensive adverse feedback index threshold set comprising a first comprehensive adverse feedback index threshold, a second comprehensive adverse feedback index threshold, and a third comprehensive adverse feedback index threshold, marking the set based on the judgment and analysis results, and generating corresponding feedback suggestions.

[0153] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0154] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring, characterized in that: The following steps are involved: Obtain usage monitoring feedback information for several age groups of the set cosmetics and input it into a pre-trained multimodal recognition model for predictive analysis, thereby obtaining a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics. The data are then analyzed separately to obtain a skin adverse feedback index and a usage adverse feedback index for each age group of the set cosmetics; The multimodal recognition model is specifically a visual-text joint model, which includes a visual encoder layer, a text encoder layer, a joint embedding layer, and a regression output layer. The specific steps of obtaining a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of a set cosmetic are as follows: In the visual encoder layer of the joint visual-text model, image feature extraction is performed on the usage monitoring feedback information of the set cosmetics for each age group to obtain a high-dimensional image feature vector for each age group of the set cosmetics; In the text encoder layer of the visual-text joint model, text feature extraction is performed on the usage monitoring feedback information of the set cosmetics for each age group to obtain the text feature vector for each age group of the set cosmetics; In the joint embedding layer of the joint visual-text model, the high-dimensional image feature vector and text feature vector of each age group of the set cosmetics are fused to obtain a joint feature vector for each age group of the set cosmetics; In the regression output layer of the visual-text joint model, regression prediction processing is performed on the joint feature vector of each age group of the set cosmetics to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; The skin condition feedback data set includes skin oil secretion level value, erythema index, skin keratin thickness value, skin surface moisture content value, skin temperature gradient value, melanin index, and skin conductivity value; The usage behavior evaluation feedback data set includes usage frequency value, usage dosage value, duration value, product mixing index, comfort index, stickiness index, skin gloss index, and stinging index; Obtain the ingredient feedback safety index of the set cosmetics, and combine it with the skin adverse feedback index of each age group and the adverse feedback index for comprehensive analysis to obtain the comprehensive adverse feedback index of each age group of the set cosmetics, and perform judgment analysis with the preset comprehensive adverse feedback index threshold set, and mark it based on the judgment analysis results, and generate corresponding feedback suggestions; The specific formula for calculating the comprehensive adverse feedback index for each age group of a given cosmetic is as follows: ； in, To set the cosmetic The comprehensive negative feedback index of each age group, 、 The first step to set cosmetics Skin adverse feedback index and usage adverse feedback index for each age group, To set the safety index of cosmetic ingredients feedback, 、 、 、 The following are the skin adjustment coefficient, usage adjustment coefficient, ingredient adjustment coefficient, and adverse interaction coefficient stored in the database. 1, 2, 3, ..., , The number of age groups.

2. The method for analyzing and feedbacking adverse reactions to cosmetics based on active monitoring according to claim 1, characterized in that: The specific steps for obtaining the skin adverse feedback index for each age group of the set cosmetics are as follows: Read the skin oil secretion level value, erythema index, skin keratin thickness value, skin surface moisture content value, skin temperature gradient value, melanin index, and skin conductivity value from the skin condition feedback data set for each age group of the set cosmetics; And obtain the reference value of skin oil secretion level, skin keratin thickness, skin surface moisture content and environmental factors for each age group for which cosmetics are set; Comprehensively analyze the reference values of skin oil secretion level, skin keratin thickness, skin surface moisture content, skin oil secretion level, erythema index, skin keratin thickness, and skin surface moisture content for each age group of the set cosmetics to obtain the skin damage index for each age group of the set cosmetics; A comprehensive analysis is conducted on the skin temperature gradient value, melanin index, skin conductivity value, and environmental factors for each age group of the set cosmetics to obtain the skin sensitivity index for each age group of the set cosmetics. A comprehensive analysis is then conducted in combination with the skin damage index to obtain the skin adverse feedback index for each age group of the set cosmetics.

3. The method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring according to claim 2, characterized in that: The specific formulas for calculating the skin damage index, skin sensitivity index, and skin adverse feedback index for each age group of cosmetics are as follows: ； in, To set the cosmetic Skin damage index for each age group, 、 、 、 、 、 、 The first step to set cosmetics Skin oil secretion level value, skin oil secretion level reference value, erythema index, skin keratin thickness value, skin keratin thickness reference value, skin surface moisture content value, skin surface moisture content reference value, 、 、 、 The following are the oil secretion adjustment coefficient, erythema adjustment coefficient, skin keratin adjustment coefficient, and moisture content adjustment coefficient stored in the database. To set the cosmetic Skin sensitivity index for each age group, 、 、 、 The first step to set cosmetics Skin temperature gradient value, melanin index, skin conductivity value, environmental factors for each age group, 、 、 、 、 They are the sensitive interaction coefficient, temperature gradient adjustment coefficient, melanin adjustment coefficient, conductivity adjustment coefficient, and environment adjustment coefficient stored in the database. To set the cosmetic Skin adverse feedback index for each age group, 、 、 、 The damage coefficient, damage adjustment coefficient, sensitivity coefficient, and sensitivity adjustment coefficient stored in the database are listed in order. , is a natural constant, 1, 2, 3, ..., , The number of age groups.

4. The method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring according to claim 1, characterized in that: The specific steps for obtaining the adverse feedback index of the cosmetics used in each age group are as follows: Read the usage frequency value, usage dosage value, duration value, product mixing index, comfort index, stickiness index, skin gloss index, and stinging index from the usage behavior evaluation feedback dataset for each age group of a given cosmetic; A comprehensive analysis is conducted on the usage frequency, dosage, duration, and product mixing index of the cosmetics for each age group to obtain the usage risk index for each age group. A comprehensive analysis is conducted on the comfort index, stickiness index, skin gloss index and stinging index of the set cosmetics for each age group to obtain the sensory evaluation index of the set cosmetics for each age group, and a comprehensive analysis is conducted in combination with the usage risk index to obtain the adverse feedback index of the set cosmetics for each age group.

5. The method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring according to claim 4, characterized in that: The specific formula for calculating the sensory evaluation index and adverse feedback index of cosmetics for each age group is as follows: ； in, To set the cosmetic Sensory evaluation index of each age group, 、 、 、 The first step to set cosmetics Comfort index, skin gloss index, stickiness index, and tingling index for each age group. 、 、 They are the positive adjustment coefficient, negative adjustment coefficient, and difference adjustment coefficient stored in the database, To set the cosmetic The negative feedback index for each age group, To set the cosmetic The risk index for use in each age group, 、 、 、 The following are the behavior adjustment coefficient, sensory evaluation adjustment coefficient, behavior and sensory evaluation difference adjustment coefficient, and behavior and sensory evaluation interaction adjustment coefficient stored in the database. 1, 2, 3, ..., , The number of age groups.

6. The method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring according to claim 1, characterized in that: The specific steps to obtain the feedback safety index of the ingredients of the set cosmetics are as follows: Obtain the concentration value, concentration reference value, irritation score value, allergenicity score value, and acnegenicity score value of each ingredient in the specified cosmetics; A comprehensive analysis is also conducted on the concentration value, concentration reference value, irritation score value, allergenicity score value, and acne-causing score value of each ingredient of the set cosmetics to obtain the ingredient feedback safety index of the set cosmetics.

7. The method for analyzing and feedback of adverse reactions to cosmetics based on active monitoring according to claim 1, characterized in that: The comprehensive negative feedback index threshold set includes a first comprehensive negative feedback index threshold, a second comprehensive negative feedback index threshold, and a third comprehensive negative feedback index threshold. The specific steps of generating corresponding feedback suggestions based on the judgment and analysis results are as follows: If the comprehensive adverse feedback index of the cosmetics for each age group is lower than or equal to the preset first comprehensive adverse feedback index threshold, it is marked as no adverse reaction and a first feedback suggestion is generated; If the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset first comprehensive adverse feedback index threshold and lower than or equal to the preset second comprehensive adverse feedback index threshold, it is marked as a mild adverse reaction and a second feedback suggestion is generated; If the comprehensive adverse feedback index of the cosmetics for each age group is higher than the preset second comprehensive adverse feedback index threshold and lower than or equal to the preset third comprehensive adverse feedback index threshold, it is marked as a moderate adverse reaction and a third feedback suggestion is generated; If the comprehensive adverse feedback index of the cosmetics for each age group is set to be higher than the preset third comprehensive adverse feedback index threshold, it will be marked as a severe adverse reaction and a fourth feedback suggestion will be generated.

8. A cosmetic adverse reaction analysis and feedback system based on active monitoring, applying the cosmetic adverse reaction analysis and feedback method based on active monitoring according to any one of claims 1 to 7, characterized in that: include: The data acquisition module is used to obtain usage monitoring feedback information of the set cosmetics for several age groups and input it into a pre-trained multimodal recognition model for predictive analysis to obtain a skin condition feedback dataset and a usage behavior evaluation feedback dataset for each age group of the set cosmetics; A data analysis module is used to analyze the skin condition feedback data set and the usage behavior evaluation feedback data set for each age group of the set cosmetics, and obtain the skin adverse feedback index and usage adverse feedback index for each age group of the set cosmetics; A comprehensive analysis module is used to simultaneously obtain the ingredient feedback safety index of the set cosmetics, and conduct a comprehensive analysis based on the skin adverse feedback index and the use adverse feedback index of each age group to obtain the comprehensive adverse feedback index of the set cosmetics for each age group; The judgment and feedback module is used to judge and analyze the comprehensive negative feedback index of each age group of the set cosmetics with a preset comprehensive negative feedback index threshold set, wherein the comprehensive negative feedback index threshold set includes a first comprehensive negative feedback index threshold, a second comprehensive negative feedback index threshold, and a third comprehensive negative feedback index threshold, and mark based on the judgment and analysis results, and generate corresponding feedback suggestions.

Images