Obesity risk subtype prediction system and method based on multi-source data

Abstract

The invention discloses an obesity risk subtype prediction method based on multi-source data, and the method comprises the steps: obtaining the multi-source data, and carrying out the standardization processing of the multi-source data; performing first model and second model training on the standardized multi-source data to generate an obesity risk level prediction model and a subtype prediction model; based on the obesity risk level prediction model and the subtype prediction model, determining a target risk level and a target subtype corresponding to multi-source data of a target user; and selecting a corresponding dietary conditioning scheme according to the target risk level and the target subtype type. The invention further provides an obesity risk subtype prediction system based on the multi-source data. According to the method, the report containing risk assessment, subtype identification and staged personalized conditioning schemes is generated through the dual-output model in combination with six obesity subtype identification standards and conditioning schemes, and the problem that obesity subtype accurate identification and targeted intervention cannot be achieved in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the intersection of precision medicine and artificial intelligence, and in particular to a system and method for predicting obesity risk subtypes based on multi-source data. Background Technology

[0002] Obesity, a heterogeneous metabolic disease caused by multiple factors, is characterized by excessive body fat. Its causes involve various factors including genetics, environment, and lifestyle. Obesity typically increases the risk of serious health problems such as heart disease, stroke, type 2 diabetes, and certain types of cancer. Clinically, Body Mass Index (BMI) is commonly used as a measurement indicator; when a BMI exceeds the threshold set by the World Health Organization or relevant national standards, obesity is diagnosed. Diet is one of the means of regulating and improving obesity prevention and treatment.

[0003] However, obesity varies from person to person and is strongly correlated with age and gender. Furthermore, using BMI or other single dimensions (such as genetics / lifestyle) to measure or assess obesity risk is highly inaccurate. In addition, dietary approaches to improve obesity often suffer from a "one-size-fits-all" approach, offering largely the same dietary plan for different obesity symptoms without providing individualized recommendations. Therefore, developing a system capable of predicting a user's obesity risk is of great significance. Summary of the Invention

[0004] Therefore, it is necessary to provide a system and method for predicting obesity risk subtypes based on multi-source data to address the aforementioned technical problems.

[0005] A method for predicting obesity risk subtypes based on multi-source data includes:

[0006] Acquire multi-source data and standardize the multi-source data, wherein the multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information;

[0007] The first model was trained on the standardized multi-source data to generate an obesity risk level prediction model.

[0008] A second model is trained on the standardized multi-source data to generate a subtype prediction model.

[0009] Based on the aforementioned obesity risk level prediction model and subtype type prediction model, the target risk level and target subtype type corresponding to the multi-source data of the target user are determined; and

[0010] Select the corresponding dietary conditioning plan based on the target risk level and target subtype.

[0011] Furthermore, the dietary habit information includes one or more first indicators, the lifestyle information includes one or more second indicators, the subtype association information includes one or more third indicators, and the medical history information includes one or more fourth indicators. The standardization processing of the multi-source data includes:

[0012] The first indicator of dietary habits information, the second indicator of lifestyle information, the third indicator of subtype association information, and the fourth indicator of medical history information are all subjected to min-max normalization to map the data of each indicator to... Within the interval, the minimum-maximum normalization formula is:

[0013]

[0014] in, The original data, and These represent the maximum and minimum values ​​of the corresponding indicator data. Normalized data;

[0015] The data of the first indicator of the dietary habit information after normalization are weighted and summed to obtain the dietary habit score;

[0016] The lifestyle information is obtained by weighted summation of the normalized data of the second indicator;

[0017] The subtype association score is obtained by weighted summation of the normalized data of the third indicator of the subtype association information.

[0018] The data of the fourth indicator of the medical history information, after normalization, are weighted and summed to obtain the medical history score.

[0019] Furthermore, the first model is a softmax regression model, and the step of training the first model on the standardized multi-source data to generate an obesity risk level prediction model includes:

[0020] A feature matrix was constructed based on dietary habit scores, lifestyle scores, subtype association scores, and medical history scores.

[0021] The feature matrix is ​​trained based on the softmax regression model to generate the obesity risk level prediction model; wherein, the softmax regression model is:

[0022]

[0023] in, This indicates three risk levels, with values ​​of 0, 1, or 2. Indicates user Classified as risk level The probability, Indicates risk level The bias term is used to adjust the base probability of the risk level. Indicates risk level The weight, Indicate category The score;

[0024] The softmax regression model uses LogisticRegression from the sklearn package, with multinomial as the multi-class category, sag as the optimization algorithm, regularization strength of 1.0, and a maximum number of iterations of 1000.

[0025] Furthermore, the second model is an XGBoost algorithm model, and the step of training the second model on the standardized multi-source data to generate a subtype prediction model includes:

[0026] The feature matrix is ​​identified based on the XGBoost algorithm model to generate the subtype prediction model. The output class of the XGBoost algorithm model is multi:softmax, the number of classes is 6, the maximum tree depth is 5, the learning rate is 0.1, and the number of trees is 200.

[0027] Furthermore, the step of determining the target risk level and target subtype type corresponding to the multi-source data of the target user based on the obesity risk level prediction model and subtype type prediction model includes:

[0028] The multi-source data of the target user is input into the obesity risk level prediction model to obtain the probability of the target user corresponding to each risk level;

[0029] The target risk value is determined based on the preset risk level weights and the probability of each risk level.

[0030] The target risk type is determined based on the target risk level and the target risk value, wherein the target risk type is one of low risk, medium risk, and high risk.

[0031] Furthermore, the target subtype is one of the following categories: edema type, glycosuria type, refractory type, stress type, localized type, and adolescent type.

[0032] Furthermore, the method also includes:

[0033] Select the corresponding dietary conditioning plan based on the target risk level and the target subtype;

[0034] An assessment report is generated based on the target risk level, target subtype, and dietary conditioning plan.

[0035] An obesity risk subtype prediction system based on multi-source data includes:

[0036] An acquisition model is used to acquire multi-source data and standardize the multi-source data, wherein the multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information;

[0037] The generative model is used to train a first model on standardized multi-source data to generate an obesity risk level prediction model; it is also used to train a second model on standardized multi-source data to generate a subtype prediction model.

[0038] The prediction model is used to determine the target risk level and target subtype type corresponding to the multi-source data of the target user based on the obesity risk level prediction model and the subtype type prediction model; and

[0039] The reporting model is used to select the corresponding dietary conditioning plan based on the target risk level and target subtype.

[0040] Furthermore, the dietary habit information includes one or more first indicators, the lifestyle information includes one or more second indicators, the subtype association information includes one or more third indicators, and the medical history information includes one or more fourth indicators. The acquisition module is also used to perform min-max normalization processing on the first indicators of the dietary habit information, the second indicators of the lifestyle information, the third indicators of the subtype association information, and the fourth indicators of the medical history information, respectively, to map the data of each indicator to... Within the interval, the minimum-maximum normalization formula is:

[0041]

[0042] in, The original data, and These represent the maximum and minimum values ​​of the corresponding indicator data. Normalized data;

[0043] The acquisition module is also used to perform weighted summation on the data of the first indicator of the dietary habit information after normalization to obtain a dietary habit score.

[0044] The acquisition module is also used to perform weighted summation on the data of the second indicator of the lifestyle information after normalization to obtain a lifestyle score;

[0045] The acquisition module is also used to perform weighted summation on the data of the third indicator of the subtype association information after normalization to obtain the subtype association score;

[0046] The acquisition module is also used to perform weighted summation on the data of the fourth indicator of the medical history information after normalization to obtain a medical history score.

[0047] An electronic device includes at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the obesity risk subtype prediction method based on multi-source data as described in any one of claims 1 to 7.

[0048] This invention provides a multi-source data-based obesity risk subtype prediction system and method. By integrating Softmax regression and XGBoost algorithm to construct a dual-output model of "risk level + subtype type", and combining it with a knowledge base containing identification criteria and treatment plans for 6 obesity subtypes (edema type, glycogen type, refractory type, stress type, localized type, and adolescent type), the system generates a report that includes risk assessment, subtype identification, and phased personalized treatment plans. This solves the problem that existing technologies cannot achieve accurate identification and targeted intervention of obesity subtypes. Attached Figure Description

[0049] Figure 1 This is a flowchart of the obesity risk subtype prediction method based on multi-source data according to the present invention.

[0050] Figure 2 This is a block diagram of the obesity risk subtype prediction system based on multi-source data of the present invention.

[0051] Figure 3 This is a schematic diagram of the structure of the electronic device involved in the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0053] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0054] Please see Figure 1 The diagram shows a flowchart of the obesity risk subtype prediction method based on multi-source data according to the present invention, which includes the following steps:

[0055] S100, acquire multi-source data and perform standardization processing on the multi-source data, wherein the multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information.

[0056] In this embodiment, obesity can be classified into six types: edema type, glycosuria type, refractory type, stress type, localized type, and adolescent type. Among them, the physical characteristics of edema type are: soft, sagging, and inelastic flesh with poor skin-flesh connection; the core causes include: long-term vegetarianism and nutritional deficiencies (protein / essential fatty acids); key physical symptoms include: 1. easy fatigue; 2. loss of appetite; 3. constipation; 4. high humidity; 5. lack of energy; 6. poor mood; 7. high blood pressure, high blood lipids, high blood sugar, and high uric acid.

[0057] The physical characteristics of a sugary body type are: bloated flesh (like a balloon), and cellulite on the thighs / arms; the core causes include: high sugar / high carbohydrate intake (staple foods / snacks / drinks); key physical symptoms include: 1. large mood swings; 2. gaining weight even from drinking water; 3. rough skin, prone to acne; 4. thick shoulders, belly, and thighs, high joint pressure, and limited movement; 5. feeling swelling in the calves in the afternoon; 6. constipation; 7. drowsiness; 8. breathing difficulties and brain hypoxia; 9. easy fatigue.

[0058] The characteristics of a stubborn body type include: uneven, loose, and flabby flesh; core causes include: repeated dieting and intestinal nerve disorders; key physical symptoms include: 1. Dieting leads to a continuous rebound in weight after intestinal flora imbalance; 2. Drug-induced weight loss: suppressing the brain's desire to eat, leading to continued weight gain through binge eating when medication is stopped; 3. Massage-induced weight loss: massage techniques help metabolism, but without external assistance, internal peristalsis is not stimulated, easily leading to fat accumulation; 4. Excessive exercise increases appetite; 5. Repeated dieting leads to increasing weight gain; 6. Constipation; 7. Fatigue; 8. Breathing obstruction, leading to cerebral hypoxia; 9. Loose flesh; 10. Binge eating.

[0059] The physical characteristics of a stress-related constitution are: bloated flesh, dry skin, and poor water absorption; the core causes include: large mood swings, overeating, and insufficient sleep (<6 hours / day); key physical symptoms include: 1. high work and life stress, poor sleep; 2. short sleep duration; 3. unstable emotions; 4. migraines; 5. dry eyes, decreased vision; 6. high blood pressure, high blood sugar, and high cholesterol; 7. hypothyroidism; 8. mental illness; 9. fatigue.

[0060] Localized body characteristics include: large belly / thick thighs, uneven muscle firmness; core causes include: prolonged sitting, high salt / heavy-flavored diet, and alcohol consumption (large belly); key physical symptoms include: 1. Large belly: due to alcohol consumption or dehydration; 2. Prolonged sitting; 3. Irregular eating habits; 4. Constipation or diarrhea, with frequent bowel movements due to excessive harmful bacteria in those who drink alcohol regularly; 5. Flatulence; 6. Leg soreness; 7. Fatigue.

[0061] The physical characteristics of adolescents are: overweight (BMI≥24), loose flesh; core causes include: high sugar junk food, lack of exercise, and irregular sleep patterns; key physical symptoms include: 1. stunted growth; 2. closure of growth plates due to overeating; 3. precocious puberty: excessive consumption of junk food, additives, and hormone-containing foods; 4. hypertension: hypertension has reached 31% in children; 5. hyperlipidemia: excessive intake of sugary drinks and meat; 6. diabetes: insulin disorder; 7. low self-esteem; 8. respiratory obstruction: cerebral hypoxia; 9. intellectual decline.

[0062] Each type of obesity is classified into three risk levels: low, medium, and high. In other implementations, the number of obesity types and / or risk levels can also be adjusted or changed accordingly.

[0063] In this embodiment, multi-source data includes, but is not limited to, dietary habit information, lifestyle information, subtype association information, disease history information, and dietary management information. The various data points from these multi-source sources can be obtained through methods such as scoring or image recognition.

[0064] For example, dietary habit information includes indicators such as the number of meals per day, the number of times one eats at home per month, preferred and disliked foods, frequency of food intake, preferred cooking methods, whether one overeats or diets, duration of vegetarianism, and types of snacks consumed. Lifestyle information includes indicators such as type and frequency of exercise, sleep duration, sleep quality, sleep quality, sleep duration, smoking status, and alcohol consumption. Because the units and numerical ranges of different indicators vary greatly—for example, the number of meals per day is measured in "times," and sleep duration in minutes—standardization is necessary to eliminate the dominant influence of large numerical features in subsequent model training.

[0065] Dietary habit and lifestyle information can be obtained through scoring. Different weights can be assigned to various indicators of dietary habit and lifestyle information. Then, the standardized data can be weighted and summed to obtain the dietary habit score and lifestyle score, respectively. This embodiment preferentially uses the minimum-maximum normalization method to linearly map the data of each indicator to... The interval is calculated using the following formula:

[0066]

[0067] in, The original data, and These are the maximum and minimum values ​​of the indicator data, respectively. Normalized data. In some scenarios, this can also be used. Standardize it so that it conforms to the mean. The standard deviation is It follows a standard normal distribution.

[0068] Subtype association information includes meat quality characteristics (such as meat firmness, expansion, and unevenness, which can be detected through image recognition, with corresponding scores of 0-1), body shape dimensions (such as height, waist circumference, thigh circumference, and calf circumference; waist / height ratio and thigh / calf circumference ratio can be used as body shape difference coefficients), mood and stress data (such as frequency of mood fluctuations and stress scores, with stress scores ranging from 0 to 10), age stratification data (adolescents and adults), and weight loss history data (such as number of weight loss attempts, weight loss methods (medication, massage, acupuncture, cupping, thread embedding, medicated pad heat therapy, instruments, and exercise) and time, frequency, and weight rebound; weight loss attempts greater than or equal to 3 are assigned a weight of 1.2-1.5). Simultaneously, the subtype association score is obtained by standardizing and weighted summing the various indicators of the subtype association information. The weighting and standardization processes are obtained in the same way as the dietary habit score.

[0069] Medical history information includes gender, age, and clinical phenotypic variables such as glycated hemoglobin (HbA1c), insulin resistance index (HOMA-IR), blood glucose, systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG), alanine aminotransferase (ALT), and creatinine (CREAT). The risk of different diseases (such as cardiovascular disease and diabetes) differs between men and women, and different scores can be assigned based on the gender specificity of the target disease. For example, for type 2 diabetes, men have a slightly higher risk than women, so a weight of 1.2 can be assigned to men and a weight of 1.0 to women (with women as the reference). Age: The risk of most chronic diseases increases with age, and linear or piecewise weighting can be used. For example, with 40 years old as the baseline (assigned a value of 1.0), the weight increases by 0.2 for every 10 years (e.g., 1.2 for 50 years old, 1.4 for 60 years old, 1.6 for 70 years old, etc.); or based on age-stratified risk of diseases (e.g., the risk of a certain disease is significantly increased in the 65+ age group), a higher weight is assigned to a specific age group.

[0070] In addition, glycated hemoglobin (HbA1c) reflects long-term blood glucose control and is directly related to diabetic complications and cardiovascular risk. Normal range (<5.7%): score 1.0; prediabetes (5.7%-6.4%): score 1.5; diabetes (≥6.5%): score 2.0 (can be further adjusted upward if other high-risk factors are present).

[0071] The Homeostasis Model of Insulin Resistance (HOMA-IR) is a core mechanism of metabolic syndrome and type 2 diabetes. Normal (<2.5): score 1.0; Mild resistance (2.5-5): score 1.5; Moderate to severe resistance (≥5): score 2.0 (or higher, as severe resistance is strongly associated with disease progression).

[0072] Blood glucose (fasting blood glucose / postprandial blood glucose): directly reflects the state of glucose metabolism. Normal (fasting <6.1mmol / L, 2h postprandial <7.8mmol / L): score 1.0; Prediabetes (fasting 6.1-6.9mmol / L, or postprandial 7.8-11.1mmol / L): score 1.5; Diabetes (fasting ≥7.0mmol / L, or postprandial ≥11.1mmol / L): score 2.0.

[0073] Systolic blood pressure (SBP) / diastolic blood pressure (DBP): Hypertension is a major risk factor for cardiovascular disease. Normal blood pressure (SBP < 120 mmHg and DBP < 80 mmHg): Score 1.0; Prehypertension (SBP 120-139 mmHg or DBP 80-89 mmHg): Score 1.3; Grade 1 hypertension (SBP 140-159 mmHg or DBP 90-99 mmHg): 1.6; Grade 2 hypertension and above (SBP ≥ 160 mmHg or DBP ≥ 100 mmHg): Score 2.0 (the higher the blood pressure, the higher the score).

[0074] Total cholesterol (TC): Elevated levels are closely associated with atherosclerosis and cardiovascular events. Ideal range (<5.2 mmol / L): Score 1.0; Borderline elevated (5.2-6.2 mmol / L): Score 1.3; Elevated (≥6.2 mmol / L): Score 1.6; Significantly elevated (≥7.2 mmol / L): Score 2.0.

[0075] High-density lipoprotein cholesterol (HDL-C): "Good cholesterol," low levels are a risk factor. Ideal (men ≥1.0 mmol / L, women ≥1.3 mmol / L): score 1.0; Lower (men <1.0 mmol / L, women <1.3 mmol / L): score 1.3 (the score gradient can be appropriately increased for each 0.2 mmol / L decrease).

[0076] Triglycerides (TG): Elevated levels are associated with metabolic disorders, pancreatitis, etc. Normal (<1.7 mmol / L): 1.0; Elevated (1.7-2.3 mmol / L): 1.3; Significantly elevated (≥2.3 mmol / L): 1.6; Very high (≥5.6 mmol / L): 2.0 (directly associated with the risk of acute pancreatitis).

[0077] Alanine aminotransferase (ALT): Reflects liver function. Elevated levels often indicate fatty liver, hepatitis, and are associated with metabolic syndrome. Normal (male <50 U / L, female <35 U / L): Score 1.0; Mildly elevated (50-100 U / L): Score 1.3; Moderately elevated (100-200 U / L): Score 1.6; Severely elevated (≥200 U / L): Score 2.0 (The score may be adjusted depending on the underlying cause, such as non-alcoholic fatty liver disease or viral hepatitis).

[0078] Creatinine (CREAT): Reflects kidney function; elevated levels indicate kidney damage and are an independent risk factor for cardiovascular disease. Normal (men 59-104 μmol / L, women 45-84 μmol / L): score 1.0; Mildly elevated (105-133 μmol / L): score 1.3; Moderately elevated (134-177 μmol / L): score 1.6; Severely elevated (≥178 μmol / L): score 2.0 (the worse the kidney function, the higher the risk score). The medical history score is obtained by standardizing and weighting the various indicators of the medical history information. The weighting and standardization processes are the same as those used to obtain the dietary habit score.

[0079] In this embodiment, based on the definitions of six obesity types, the dietary habits, lifestyle, subtype association, and medical history information of each user (or sample) are determined to correspond to a specific obesity type; furthermore, based on three risk types, the specific risk level corresponding to the dietary habits, lifestyle, subtype association, and medical history information of each user (or sample) is determined.

[0080] In this embodiment, the system also includes dietary conditioning information, which stores dietary conditioning plans for different types of obesity and risk levels. The data comes from evidence-based medicine guidelines, clinical research, and practical verification. Each dietary conditioning plan may include several stages, such as a stage one conditioning plan, a stage two conditioning plan, a stage three conditioning plan, and a stage four conditioning plan. The stage one conditioning plan includes the following recipes: 1-100g kiwi, 1-100g blueberries, a microbial flora conditioning active enzyme composition, 1-100g broccoli, 1-300ml green tea, 1-100g grapefruit, 1-100g lemon, and 1-100g romaine lettuce; the conditioning period is one course of treatment, 20 days. The stage two conditioning plan includes appropriate amounts of cucumber juice, cabbage juice, lily juice, a microbial flora conditioning composition, and at least 2000ml of water; the conditioning period is 9 days. The three-stage conditioning program includes a meal plan for breakfast, lunch, and dinner, with each meal potentially served at specific times. The diet includes fats, proteins, vegetables, carbohydrates, fruits, nuts, eggs, and dairy products. The conditioning cycle is one 20-day course. The four-stage conditioning program is essentially the same as the three-stage program, except for the conditioning cycle. The four-stage program lasts for three months, allowing users to achieve comprehensive cellular optimization, a stable and balanced gut microbiota, and a lean physique throughout the four stages.

[0081] In this embodiment, data such as dietary habits, lifestyle, subtype association, medical history, and dietary management information for different subtypes can be collected or acquired before, during, and / or after the start of a given dietary management program cycle. The dietary management information stores effective plans for different obesity types. That is, after the dietary management program cycle ends, all subtype-related indicators for different obesity types are normal or the child's weight reaches the standard weight. For example, for glycosuria, after the dietary management program cycle ends, all subtype-related indicators are normal, mood is stable, and weight reaches the standard weight. For adolescent obesity, gut microbiota is balanced, and problems related to the child's skin, bones, height, and concentration are resolved.

[0082] S200 uses standardized multi-source data to train the first model and generate an obesity risk level prediction model.

[0083] In this embodiment, standardized multi-source data is input into the first model for training to obtain an obesity risk level prediction model. The obesity risk level prediction model outputs the probability that a user belongs to high-risk, medium-risk, or low-risk categories. A feature matrix is ​​constructed based on dietary habit scores, lifestyle scores, subtype association scores, and medical history scores, and used as input to the first model to generate the obesity risk level prediction model. The first model is a softmax regression model, a classic multi-classification algorithm. Its core function is to predict the probability of a sample belonging to each category based on the input features (normalized using the softmax function), and finally output the probability distribution of the specific categories. The softmax regression model is as follows:

[0084]

[0085] in, This indicates three risk levels, with values ​​of 0, 1, or 2. Indicates user Classified as risk level The probability, Indicates risk level The bias term is used to adjust the base probability of the risk level. Indicates risk level The weight, Indicate category The score. Weights are adjusted using optimization algorithms (such as gradient descent). With bias This minimizes the cross-entropy loss between the predicted probability and the true risk level label. The relevant parameters for model training are as follows:

[0086] LogisticRegression(

[0087] multi_class='multinomial', # Multiclassification

[0088] solver='sag', # Optimization algorithm

[0089] C=1.0, # Regularization strength

[0090] max_iter=1000, # Maximum number of iterations

[0091] random_state=42 # Random number seed )

[0093] The first model uses LogisticRegression from the sklearn package. Its parameters are as follows: multi_class is multinomial, solver is sag, regularization strength C is 1.0, maximum number of iterations max_iter is 1000, and random_state is 42.

[0094] S300 trains a second model on standardized multi-source data to generate a subtype prediction model.

[0095] In this embodiment, standardized multi-source data is input into the second model for training to obtain the subtype prediction model. Specifically, a feature matrix is ​​constructed based on dietary habit scores, lifestyle scores, subtype association scores, and medical history scores, and this matrix is ​​used as input to the second model to generate the subtype prediction model. The subtype prediction model can output the user's category as edema-type, glycosuria-type, refractory, stress-type, localized, or adolescent-type. The second model is an XGBoost algorithm model. The training parameters of the XGBoost algorithm model are as follows:

[0096] XGBClassifier

[0097] objective='multi:softmax', # Output category

[0098] num_class=6, # 6 types of obesity

[0099] max_depth=5, # Maximum depth of the tree

[0100] learning_rate=0.1, # Learning rate

[0101] n_estimators=200, # Number of trees

[0102] subsample=0.8, # Use 80% of the samples for each training iteration

[0103] random_state=42 # Fixed random number )

[0105] The second model uses XGBClassifier from the xgboost package. Its parameters are as follows: output class objective is multi:softmax; number of classes num_class is 6 (corresponding to edema type, glycogen type, refractory type, stress type, local type, or adolescent type); maximum tree depth max_depth is 5; learning rate learning_rate is 0.1; number of trees n_estimators is 200; training sample rate subsample is 0.8; and seed random number random_state is 42.

[0106] S400, based on the obesity risk level prediction model and subtype type prediction model, determine the target risk level and target subtype type corresponding to the multi-source data of the target user.

[0107] In this embodiment, after inputting multi-source data for the target user into the obesity risk level prediction model, the probability of the target user corresponding to each risk level is obtained. Based on the preset risk level weights and the probability of each risk level, the target risk value is determined. And based on the target risk level and the target risk value, the target risk type is determined. For example, after inputting the target user's dietary habits, lifestyle information, subtype association information, and medical history information into the obesity risk level prediction model, the probability P1 of low risk, the probability P2 of medium risk, and the probability P3 of high risk for the target user are obtained, where P1 + P2 + P3 = 1. The target risk value = P1 * W1 + P2 * W2 + P3 * W3, where W1, W2, and W3 are risk level weights, and the sum of the three is 1. After obtaining the target risk value, the target risk type is determined based on the target risk level. For example, when the target risk value is less than the first value, the target risk type can be determined as low risk; when the target risk value is not less than the first value and less than the second value, the target risk type can be determined as medium risk; when the target risk value is not less than the second value, the target risk type can be determined as high risk.

[0108] After inputting multi-source data for the target user into the subtype prediction model, the target subtype type of the target user is obtained. For example, after inputting the target user's dietary habits, lifestyle information, subtype association information, and medical history information into the subtype prediction model, the target user is identified as one of the following categories: edema type, diabetes type, refractory type, stress type, localized type, and adolescent type.

[0109] In other implementations, both the obesity risk level prediction model and the subtype type prediction model can employ incremental learning techniques to continuously incorporate new obesity label data to update the model and optimize the feature weights in the model.

[0110] S500 selects the corresponding dietary conditioning plan based on the target risk level and target subtype.

[0111] In this embodiment, after determining the target user's target risk level and target subtype, a corresponding dietary management plan can be selected from the dietary management information. For example, if the target user's risk level is determined to be high-risk and the target subtype to be refractory, the first dietary management plan is selected; if the target user's risk level is determined to be medium-risk and the target subtype to be refractory, the second dietary management plan is selected. For the medium-risk edema type, the corresponding plan is as follows:

[0112] The first phase of the treatment plan includes: taking the gut microbiota conditioning juice 0.5-1 hour before meals, with a course of treatment lasting 20 days. Maintain your original dietary habits and gradually supplement with beef / eggs on a voluntary basis.

[0113] Two-stage conditioning plan: 1-5 days of fasting + 1-4 days of refeeding;

[0114] Three-stage conditioning plan: high-fiber vegetables + olive oil + protein + high-quality low-GI carbohydrates, daily water intake 2000mL).

[0115] The four-stage conditioning plan includes a three-month consolidation phase, following the dietary intake guidelines of the three-stage plan.

[0116] Furthermore, dietary plans are not recommended for adolescent obesity.

[0117] In addition, an assessment report can be generated based on the target risk level, target subtype, and dietary treatment plan. The assessment report includes:

[0118] 1) Clearly identify the subtype and explain the identification criteria for "physical characteristics + dietary habits + physical symptoms" (e.g., "You belong to edema-type obesity, based on: soft and sagging flesh (code 2 points) + vegetarianism and picky eating + > the following 3 items (1. easy fatigue; 2. no desire to eat meat; 3. constipation; 4. high humidity; 5. lack of energy; 6. poor mood; 7. high blood pressure, high blood lipids, high blood sugar, high uric acid").

[0119] 2) Risk-subtype association analysis: Explain the association between risk level and subtype (e.g., "High risk stems from nutritional deficiencies caused by edema, which easily leads to anemia").

[0120] 3) Evidence-based references: Each key conclusion and protocol is marked with its source (e.g., “The intermittent fasting protocol cites the European Guidelines for the Management of Obesity 2024”).

[0121] This invention constructs a dual-output model of "risk level + subtype type" by integrating Softmax regression and XGBoost algorithm. Combined with a knowledge base containing identification criteria and treatment plans for 6 obesity subtypes (edema type, glycogen type, refractory type, stress type, localized type, and adolescent type), it generates a report that includes risk assessment, subtype identification, and phased personalized treatment plans, solving the problem that existing technologies cannot achieve accurate identification and targeted intervention of obesity subtypes.

[0122] Please see Figure 2 The diagram shows a block diagram of the obesity risk subtype prediction system based on multi-source data according to the present invention. The obesity risk subtype prediction system 600 includes:

[0123] The acquisition module 602 is used for multi-source data and performs standardization processing on the multi-source data. The multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information. The main execution step S100 function is used.

[0124] The generation module 604 is used to train the first and second models on the standardized multi-source data to generate an obesity risk level prediction model and a subtype type prediction model. It mainly executes the functions of steps S200 and S300.

[0125] The prediction module 606 is used to determine the target risk level and target subtype type corresponding to the multi-source data of the target user based on the obesity risk level prediction model and subtype type prediction model, and mainly performs the function of step S400.

[0126] Report model 608 is used to select the corresponding dietary conditioning plan based on the target risk level and target subtype, and generate an assessment report, mainly performing the function of step S500. The obesity risk subtype prediction system in this embodiment refers to the technical solution of the aforementioned method, and will not be described in detail here.

[0127] Please see Figure 3 This invention provides an electronic device. Specifically, the electronic device 700 includes a processor 701, a memory 702, a communication interface 703, and a bus 704; wherein the processor 701, the memory 702, and the communication interface 703 can be connected to each other via the bus 704, or can be connected using other connection methods besides the bus 704.

[0128] Processor 701 may be a general-purpose processor, which can be a processor that performs specific steps and / or operations by reading and executing contents stored in memory (e.g., memory 702). For example, a general-purpose processor may be a central processing unit (CPU). Processor 701 may include at least one circuit to perform... Figure 1 All or part of the steps of the method shown.

[0129] The memory 702 can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), flash memory, optical storage, hard disk, etc.

[0130] Specifically, memory 702 can be used to store several program codes. When processor 701 executes this program code, the above-mentioned... Figure 1 The corresponding process steps.

[0131] The communication interface 703 includes input / output (I / O) interfaces, physical interfaces, and logical interfaces for interconnecting devices within the electronic device 700, as well as interfaces for interconnecting the electronic device 700 with other devices (such as other computing devices or user equipment). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.

[0132] Bus 704 can be any type of communication bus used to interconnect processor 701, memory 702 and communication interface 703, such as system bus.

[0133] The aforementioned electronic devices 700 can be disposed on separate chips, or at least partially or entirely on the same chip. Whether the devices are disposed independently on different chips or integrated on one or more chips often depends on the needs of the product design. In this invention, the specific implementation of the aforementioned electronic devices is not limited.

[0134] It needs to be explained that, Figure 3 The electronic device 700 shown is merely an example. In the implementation process, the electronic device 700 may also include other components, which will not be listed one by one in this article.

[0135] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0136] It is understood that the various numerical designations used in the embodiments of the present invention are merely for descriptive convenience and are not intended to limit the scope of the embodiments of the present invention. It should be understood that in the embodiments of the present invention, the order of the process numbers does not imply the order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0137] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting obesity risk subtypes based on multi-source data, comprising: Acquire multi-source data and standardize the multi-source data, wherein the multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information; The first model was trained on the standardized multi-source data to generate an obesity risk level prediction model. A second model is trained on the standardized multi-source data to generate a subtype prediction model. Based on the aforementioned obesity risk level prediction model and subtype type prediction model, the target risk level and target subtype type corresponding to the multi-source data of the target user are determined; and Select the corresponding dietary conditioning plan based on the target risk level and target subtype.

2. The method for predicting obesity risk subtypes as described in claim 1, characterized in that, The dietary habit information includes one or more first indicators, the lifestyle information includes one or more second indicators, the subtype association information includes one or more third indicators, and the medical history information includes one or more fourth indicators. The standardization processing of the multi-source data includes: The first indicator of dietary habits information, the second indicator of lifestyle information, the third indicator of subtype association information, and the fourth indicator of medical history information are all subjected to min-max normalization to map the data of each indicator to... Within the interval, the minimum-maximum normalization formula is: in, The original data, and These represent the maximum and minimum values ​​of the corresponding indicator data. Normalized data; The data of the first indicator of the dietary habit information after normalization are weighted and summed to obtain the dietary habit score; The lifestyle information is obtained by weighted summation of the normalized data of the second indicator; The subtype association score is obtained by weighted summation of the normalized data of the third indicator of the subtype association information. The data of the fourth indicator of the medical history information, after normalization, are weighted and summed to obtain the medical history score.

3. The method for predicting obesity risk subtypes as described in claim 2, characterized in that, The first model is a softmax regression model. The step of training the first model on standardized multi-source data to generate an obesity risk level prediction model includes: A feature matrix was constructed based on dietary habit scores, lifestyle scores, subtype association scores, and medical history scores. The feature matrix is ​​trained based on the softmax regression model to generate the obesity risk level prediction model; wherein, the softmax regression model is: in, This indicates three risk levels, with values ​​of 0, 1, or 2. Indicates user Classified as risk level The probability, Indicates risk level The bias term is used to adjust the base probability of the risk level. Indicates risk level The weight, Indicate category The score; The softmax regression model uses LogisticRegression from the sklearn package, with multinomial as the multi-class category, sag as the optimization algorithm, regularization strength of 1.0, and a maximum number of iterations of 1000.

4. The method for predicting obesity risk subtypes as described in claim 3, characterized in that, The second model is the XGBoost algorithm model. The step of training the second model on the standardized multi-source data to generate a subtype prediction model includes: The feature matrix is ​​identified based on the XGBoost algorithm model to generate the subtype prediction model. The output class of the XGBoost algorithm model is multi:softmax, the number of classes is 6, the maximum tree depth is 5, the learning rate is 0.1, and the number of trees is 200.

5. The method for predicting obesity risk subtypes as described in claim 4, characterized in that, The process of determining the target risk level and target subtype type corresponding to the multi-source data of the target user based on the obesity risk level prediction model and subtype type prediction model includes: The multi-source data of the target user is input into the obesity risk level prediction model to obtain the probability of the target user corresponding to each risk level; The target risk value is determined based on the preset risk level weights and the probability of each risk level. The target risk type is determined based on the target risk level and the target risk value, wherein the target risk type is one of low risk, medium risk, and high risk.

6. The method for predicting obesity risk subtypes as described in claim 5, characterized in that, The target subtype is one of the following categories: edema type, glycosuria type, refractory type, stress type, localized type, and adolescent type.

7. The method for predicting obesity risk subtypes as described in claim 6, characterized in that, The method further includes: Select the corresponding dietary conditioning plan based on the target risk level and the target subtype; An assessment report is generated based on the target risk level, target subtype, and dietary conditioning plan.

8. A system for predicting obesity risk subtypes based on multi-source data, comprising: An acquisition model is used to acquire multi-source data and standardize the multi-source data, wherein the multi-source data includes at least dietary habit information, lifestyle information, subtype association information, and disease history information; The generative model is used to train a first model on standardized multi-source data to generate an obesity risk level prediction model; it is also used to train a second model on standardized multi-source data to generate a subtype prediction model. The prediction model is used to determine the target risk level and target subtype type corresponding to the multi-source data of the target user based on the obesity risk level prediction model and the subtype type prediction model; and The reporting model is used to select the corresponding dietary conditioning plan based on the target risk level and target subtype.

9. The obesity risk subtype prediction system as described in claim 8, characterized in that, The dietary habit information includes one or more first indicators, the lifestyle information includes one or more second indicators, the subtype association information includes one or more third indicators, and the medical history information includes one or more fourth indicators. The acquisition module is further configured to perform min-max normalization processing on the first indicators of the dietary habit information, the second indicators of the lifestyle information, the third indicators of the subtype association information, and the fourth indicators of the medical history information, respectively, to map the data of each indicator to... Within the interval, the minimum-maximum normalization formula is: in, The original data, and These represent the maximum and minimum values ​​of the corresponding indicator data. Normalized data; The acquisition module is also used to perform weighted summation on the data of the first indicator of the dietary habit information after normalization to obtain a dietary habit score. The acquisition module is also used to perform weighted summation on the data of the second indicator of the lifestyle information after normalization to obtain a lifestyle score; The acquisition module is also used to perform weighted summation on the data of the third indicator of the subtype association information after normalization to obtain the subtype association score; The acquisition module is also used to perform weighted summation on the data of the fourth indicator of the medical history information after normalization to obtain a medical history score.

10. An electronic device, comprising at least one processor and a memory communicatively connected to said at least one processor; characterized in that, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the obesity risk subtype prediction method based on multi-source data as described in any one of claims 1 to 7.

Images