Method, device and medium for constructing obesity genetic risk model based on stacking model

Abstract

The present application provides a method, electronic device, and storage medium for constructing an obesity genetic risk model based on a stacking model. The method comprises: obtaining a data set, wherein the data set comprises genetic feature data, wherein the genetic feature data comprises single nucleotide polymorphisms; applying a random forest to screen the single nucleotide polymorphisms to obtain a first genetic feature set related to the genetic risk of obesity; applying a distributed gradient boosting library model to screen the single nucleotide polymorphisms to obtain a second genetic feature set related to the genetic risk of obesity; obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set; constructing an obesity genetic risk model; obtaining a model evaluation result of the obesity genetic risk model, and optimizing the parameters of the obesity genetic risk model based on the model evaluation result. By integrating multiple basic learners and combining meta-classifier logistic regression for stacking training, the advantages of different algorithms are effectively integrated, thereby improving the accuracy and stability of obesity genetic risk prediction.

Description

Technical Field

[0001] The present application relates to the technical field of obesity genetic risk prediction, and in particular to a method for constructing an obesity genetic risk model based on a stacking model, an electronic device, and a storage medium. Background Art

[0002] Obesity is a major public health issue worldwide and has been shown to be closely associated with a variety of chronic diseases (such as cardiovascular disease, diabetes, and metabolic syndrome). In recent years, with the development of genomics technology, the genetic basis of obesity has gradually attracted attention. Studies have shown that an individual's susceptibility to obesity is affected by genetic polymorphisms, which usually exist in the form of single nucleotide polymorphisms (SNPs). Although there have been a large number of genetic-related obesity studies, how to use this genetic information to accurately assess an individual's obesity risk remains a major challenge in current scientific research and clinical practice.

[0003] Currently, there is a lack of reliable tools in relevant technologies for assessing genetic risk of obesity and assessing the risk of death related to obesity. Summary of the Invention

[0004] In view of this, it is necessary to provide a method for constructing an obesity genetic risk model based on a stacking model, an electronic device, and a storage medium that can overcome at least one of the above defects.

[0005] In a first aspect, an embodiment of the present application provides a method for constructing an obesity genetic risk model based on a stacking model, the method comprising:

[0006] Acquiring a data set, the data set comprising genetic characteristic data, the genetic characteristic data comprising single nucleotide polymorphisms;

[0007] Random forests were applied to screen single nucleotide polymorphisms to obtain the first set of genetic features associated with the genetic risk of obesity;

[0008] A distributed gradient boosting library model was applied to screen single nucleotide polymorphisms to obtain a second set of genetic features associated with obesity genetic risk;

[0009] Obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set;

[0010] Constructing an obesity genetic risk model, wherein the obesity genetic risk model is a stacking model, and the obesity genetic risk model includes a basic learner and a meta-classifier;

[0011] The basic learners include: random forest, distributed gradient boosting library model, adaptive boosting and logistic regression;

[0012] The algorithms in the basic learner predict the genetic risk of obesity according to the significant genetic feature set, and output the prediction results to the meta-classifier respectively, and the meta-classifier is used to output the genetic risk prediction results of obesity;

[0013] Obtaining a model evaluation result of the obesity genetic risk model, and optimizing the parameters of the obesity genetic risk model according to the model evaluation result.

[0014] According to one embodiment of the present application, obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set includes:

[0015] The intersection of the first genetic feature set and the second genetic feature set is taken to obtain the significant genetic feature set.

[0016] According to one embodiment of the present application, obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set further includes:

[0017] The first genetic feature set and the second genetic feature set are analyzed according to lasso regression to obtain the significant genetic feature set.

[0018] According to one embodiment of the present application, the meta-classifier is a logistic regression model, and the meta-classifier generates a prediction value of obesity genetic risk by integrating the prediction results of the basic learner.

[0019] According to one embodiment of the present application, the first genetic feature set obtains the importance score of the genetic feature through a random forest algorithm, and the genetic features with the importance score higher than a preset threshold are selected as the first genetic feature set;

[0020] The second genetic feature set obtains the importance scores of the genetic features through a distributed gradient boosting library model, and the genetic features with importance scores higher than a preset threshold are screened as the second genetic feature set.

[0021] According to one embodiment of the present application, the method further includes:

[0022] Randomly dividing the significant genetic feature set into five subsets with equal data amounts;

[0023] Sequentially applying the four subsets to train the basic learner, and applying one subset to validate the obesity genetic risk model to form a cross-validation;

[0024] After completing five cross-validations, five sets of prediction values ​​are obtained;

[0025] A training set is formed based on the five sets of predicted values.

[0026] According to one embodiment of the present application, the data set includes an internal test queue and an external test queue, and the method further includes:

[0027] Inputting the internal test queue into each of the basic learners to obtain five sets of internal prediction values, and taking the average of the internal prediction values ​​as the internal test set;

[0028] Inputting the external test queue into each of the basic learners to obtain five sets of external prediction values, and taking the average of the external prediction values ​​as the external test set;

[0029] The obesity genetic risk prediction result is output based on the external test set input and the distributed gradient boosting library model.

[0030] According to one embodiment of the present application, obtaining the model evaluation result of the obesity genetic risk model includes:

[0031] Evaluating a model evaluation result of the obesity genetic risk model according to a model evaluation method;

[0032] The model evaluation method includes calculating the accuracy, recall rate, F1 score and area under the ROC curve of the obesity genetic risk model.

[0033] In a second aspect, an embodiment of the present application provides an electronic device comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to implement the method for constructing an obesity genetic risk model based on a stacking model as described in the first aspect when executing the instructions.

[0034] In a third aspect, an embodiment of the present application provides a computer-readable storage medium comprising instructions, wherein the instructions instruct a device to execute the method for constructing an obesity genetic risk model based on a stacking model as described in the first aspect.

[0035] The stacking model-based obesity genetic risk model construction method, electronic device, and storage medium provided in the present application effectively combines the advantages of different algorithms by integrating a basic learner and combining it with a meta-classifier logistic regression for stacking training, thereby improving the accuracy and stability of obesity genetic risk prediction. Multiple screening of single nucleotide polymorphism data and the intersection of these data to obtain a set of significant genetic features ensure that the selected features are strongly correlated with obesity genetic risk, further improving the model's feature extraction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic flow chart of a method for constructing an obesity genetic risk model based on a stacking model provided in one embodiment of the present application.

[0037] Figure 2A 、 Figure 2B 、 Figure 2C 、 Figure 2D and Figure 2E Schematic diagram of the mean square error of the obesity genetic risk model provided in the examples of this application.

[0038] Figure 3 A schematic diagram of an obesity genetic risk model provided in one embodiment of the present application.

[0039] Figure 4A and Figure 4B A schematic diagram of the obesity genetic risk model provided in one embodiment of the present application for predicting the cumulative risk of death.

[0040] Figure 5 A schematic diagram of an electronic device provided in one embodiment of the present application.

[0041] Description of main component symbols

[0042] Electronic devices 20

[0043] Processor 21

[0044] Memory 22

[0045] Method steps S100-S600 DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0047] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art in the art to which this application relates. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0048] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying an order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0049] Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] Obesity is a major public health issue worldwide and has been shown to be closely associated with a variety of chronic diseases (such as cardiovascular disease, diabetes, and metabolic syndrome). In recent years, with the development of genomics technology, the genetic basis of obesity has gradually attracted attention. Studies have shown that an individual's susceptibility to obesity is affected by genetic polymorphisms, which usually exist in the form of single nucleotide polymorphisms (SNPs). Although there have been a large number of genetic-related obesity studies, how to use this genetic information to accurately assess an individual's obesity risk remains a major challenge in current scientific research and clinical practice.

[0051] Currently, there is a lack of reliable tools in relevant technologies for assessing genetic risk of obesity and assessing the risk of death related to obesity.

[0052] In view of this, the embodiment of the present application provides a method for constructing an obesity genetic risk model based on a stacking model, an electronic device and a storage medium. By integrating a variety of basic learners such as random forest, distributed gradient boosting library model, adaptive boosting and logistic regression, and combining meta-classifier logistic regression for stacking training, the advantages of different algorithms are effectively integrated, and the accuracy and stability of obesity genetic risk prediction are improved. Single nucleotide polymorphism data are multiplexed through random forest and distributed gradient boosting library models, and the intersection is taken to obtain a set of significant genetic features to ensure that the screened features are highly correlated with obesity genetic risk, further improving the feature extraction efficiency of the model. Five cross-validations are introduced during the training process of the first layer of basic learners to avoid overfitting problems, improve the generalization ability of the model on different data sets, and improve the stability and reliability of the model. Through multi-layer training of the stacking model, the prediction results of the basic learners provide richer training data for the second layer XGBoost model. After hierarchical optimization, the model performance is more robust, enhancing its ability to predict obesity genetic risk. The model is evaluated using indicators such as accuracy, recall, F1 score, and area under the ROC curve, and the model parameters are optimized based on the evaluation results to ensure that the model can maintain good performance in practical applications.

[0053] The following describes some embodiments of the application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments may be combined with each other.

[0054] Figure 1 This is a flow chart of a method provided by an embodiment of the present application. Figure 1 The method for constructing an obesity genetic risk model based on a stacking model includes at least the following steps: S100: obtaining a data set, the data set including genetic characteristic data, the genetic characteristic data including single nucleotide polymorphisms;

[0055] S200: Apply random forest to screen single nucleotide polymorphisms to obtain a first genetic feature set related to the genetic risk of obesity; S300: Apply distributed gradient boosting library model to screen single nucleotide polymorphisms to obtain a second genetic feature set related to the genetic risk of obesity; S400: Obtain a significant genetic feature set based on the first genetic feature set and the second genetic feature set; S500: Construct an obesity genetic risk model; S600: Obtain model evaluation results of the obesity genetic risk model, and optimize the parameters of the obesity genetic risk model based on the model evaluation results.

[0056] S100: Acquire a data set, where the data set includes genetic characteristic data, and the genetic characteristic data includes single nucleotide polymorphisms.

[0057] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes obtaining a data set in step S100, where the data set includes genetic characteristic data, and the genetic characteristic data includes single nucleotide polymorphisms.

[0058] It is understandable that in the process of implementing this application, obtaining a data set is a key basic step. The data set not only contains physical examination data, but also includes genetic characteristic data, especially single nucleotide polymorphism (SNP) information related to obesity. These SNPs are the core features of studying the genetic risk of obesity and can reflect the potential relationship between gene variation and obesity. By collecting comprehensive and accurate physical examination data and SNP data, the accuracy of subsequent screening steps and the reliability of model prediction results can be ensured.

[0059] In the examples of the present application, data was first obtained from the datasets of the UK Biobank and Southern Hospital. Specifically, the data was cleaned and samples with missing BMI values ​​and genetic data were removed to ensure the integrity and reliability of the data. After data cleaning, a training cohort, an internal test cohort, and an external test cohort were obtained. The training cohort is used for model training and contains most of the data. The internal test cohort is used for internal validation of the model to evaluate the performance of the model on the training data. The external test cohort is used for independent evaluation of the model to ensure the generalization ability of the model. Through this step, the high quality of the subsequent model training data is ensured, and the influence of noise and missing data on the model results is avoided.

[0060] S200: Application of random forests to screen single nucleotide polymorphisms to obtain the first set of genetic features associated with genetic risk of obesity.

[0061] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes, in step S200 , applying a random forest to screen single nucleotide polymorphisms to obtain a first genetic feature set related to obesity genetic risk.

[0062] It is understood that random forest is an ensemble learning method that analyzes the importance of features by constructing multiple decision trees. In this embodiment, random forest is used to screen SNPs associated with obesity. The principle is to evaluate the contribution of each SNP to obesity risk prediction by calculating the importance of the split nodes in different trees. The first set of genetic features screened includes those SNPs with high importance scores in obesity risk prediction. The purpose of this step is to reduce the interference of noisy data and improve model performance.

[0063] S300: Applying a distributed gradient boosting library model to screen single nucleotide polymorphisms to obtain a second set of genetic features associated with obesity genetic risk.

[0064] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes, in step S300, applying a distributed gradient boosting library model to screen single nucleotide polymorphisms to obtain a second genetic feature set related to obesity genetic risk.

[0065] It is understandable that the distributed gradient boosting library model (such as XGBoost) is a commonly used boosting method, which is used to further screen SNPs in this embodiment. The model trains weak learners through iterative weighting, gradually reducing errors, thereby improving the effects of classification and regression. XGBoost evaluates the contribution of each SNP and screens out a second set of genetic features to ensure that these features can better reflect the risk of obesity. Compared to random forests, XGBoost can capture more complex feature relationships, so the combination of the two can fully extract SNP information.

[0066] S400: Obtain a significant genetic feature set based on the first genetic feature set and the second genetic feature set.

[0067] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes, in step S400 , obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set.

[0068] In an embodiment of the present application, obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set includes: taking the intersection of the first genetic feature set and the second genetic feature set to obtain the significant genetic feature set.

[0069] As you can understand, the set of significant genetic features is calculated by intersecting the first and second genetic feature sets. This approach ensures that the features ultimately retained are not only significant in the random forest model but also highly relevant in the XGBoost model, thereby improving the accuracy of feature selection. This method can effectively reduce redundant features and focus on extracting genetic features that are highly correlated with obesity.

[0070] In an embodiment of the present application, obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set further includes: analyzing the first genetic feature set and the second genetic feature set according to lasso regression to obtain the significant genetic feature set.

[0071] It is understood that Lasso regression (LASSO) is a regression method with L1 regularization that is used to further screen features. In this embodiment, Lasso regression introduces a penalty term to force irrelevant feature coefficients toward zero, thereby reducing model complexity and overfitting. Through Lasso regression analysis, researchers can further optimize the set of significant genetic features to ensure that the selected features have a significant contribution to obesity risk in the actual model.

[0072] In this example, random forests and XGBoost were used to perform a preliminary screening of single nucleotide polymorphisms (SNPs) recorded in the sample to identify candidate genetic features associated with obesity risk. Subsequently, LASSO regression was used to further refine these initial screening results and select the SNP features most significantly associated with obesity risk. LASSO regression, by applying L1 regularization, can avoid overfitting and retain the features most associated with obesity risk.

[0073] Please also refer to Figures 2A to 2E , Figures 2A to 2E Schematic diagram of the mean square error of the obesity genetic risk model provided in the application example.

[0074] In the examples of the present application, the embodiments of the present invention provide a demonstration of the effect of the LASSO model on screening different numbers of single nucleotide polymorphisms (SNPs) in a training set. Specifically, the figure reveals the mean square error (MSE) of the model as the number of SNPs changes. In order to determine the optimal number of SNPs, the principle of minimum mean square error is adopted, that is, the number of SNPs that minimizes the model prediction error is selected. The λ value is the regularization parameter of the LASSO model, which is used to control the complexity of the model and constrain the coefficient size of the model by introducing the L1 regularization term.

[0075] According to the diagram, as the λ value changes, the mean square error of the model shows a trend of first decreasing and then stabilizing. Finally, at the optimal λ value, i.e. 1378, a dotted line is drawn, indicating that the corresponding number of SNPs at this time can allow the model to achieve the minimum mean square error, indicating that the model at this time has the best predictive ability. This application aims to reduce the complexity of the model and prevent overfitting by adjusting the regularization parameter λ of the LASSO model, while screening out the SNPs that are most relevant to the genetic risk of obesity.

[0076] It is understandable that the embodiments of the present application can obtain significant genetic feature sets in two different ways. The first method is to take the intersection of the feature sets screened by the random forest and distributed gradient boosting library models to obtain those features that are highly important in both models. The second method is to perform further analysis through lasso regression, which uses L1 regularization to reduce the number of features and return redundant or unimportant feature coefficients to zero, ensuring that the significant genetic feature set finally screened has the best predictive effect on obesity risk in the actual model.

[0077] S500: Constructing a genetic risk model for obesity.

[0078] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes constructing an obesity genetic risk model in step S500. The obesity genetic risk model is a stacking model, and the obesity genetic risk model includes a basic learner and a meta-classifier. The basic learners include: random forest, distributed gradient boosting library model, adaptive boosting and logistic regression. Among them, the algorithms in the basic learners predict obesity genetic risk based on the significant genetic feature set, and output the prediction results to the meta-classifier respectively, and the meta-classifier is used to output the obesity genetic risk prediction results

[0079] It is understood that the obesity genetic risk model is constructed using a stacked model approach. It is understood that the stacked model is an ensemble learning method that combines multiple different base learners—random forest, distributed gradient boosting library model, adaptive boosting, and logistic regression—to enhance the model's predictive capabilities. Each base learner predicts obesity risk based on a set of significant genetic features and passes its output to a meta-classifier. The meta-classifier, lasso logistic regression, synthesizes these predictions to generate the final obesity genetic risk prediction. This model architecture helps combine the strengths of multiple learners, mitigate the limitations of a single model, and improve the accuracy and robustness of predictions.

[0080] Please also refer to Figure 3 , Figure 3 This is a schematic diagram of the obesity genetic risk model provided in one embodiment of the present application.

[0081] like Figure 3 As shown, the obesity genetic risk model takes genetic profile data as input. After passing through the basic learner's random forest, distributed gradient boosting library model, adaptive boosting, and logistic regression algorithms, it is input into the meta-classifier distributed gradient boosting library, and finally outputs the obesity-related polygenic risk score (OPRS). The OPRS value ranges from 0 to 1, with 0 indicating the lowest genetic risk for obesity and 1 indicating the highest genetic risk for obesity.

[0082] In the examples of the present application, the OPRS score uses a quartile approach, with the low-risk group being the 1st quartile, the medium-risk group being the 2nd to 3rd quartile, and the high-risk group being the 4th quartile. The 1st quartile is the low-risk group, and these participants have the lowest genetic risk for obesity. The 2nd to 3rd quartile is the medium-risk group, and these participants have a moderate genetic risk. The 4th quartile is the high-risk group, and these participants have the highest genetic risk for obesity.

[0083] Please also refer to Figure 4A and Figure 4B , Figure 4A and Figure 4B This is a schematic diagram of the obesity genetic risk model provided in one embodiment of the present application for predicting the cumulative risk of death.

[0084] like Figure 4A and Figure 4B As shown, at 5 years, the cumulative incidence of death in the high-risk group was higher than that in the low-risk and intermediate-risk groups, regardless of whether the internal training cohort or the external training cohort was included. At this time, the difference in mortality risk between the low-risk and intermediate-risk groups was not significant, indicating that the mortality rates of the low-risk and intermediate-risk groups were relatively similar in the early stages.

[0085] At 10 years, the cumulative incidence of death in the high-risk group remained significantly higher than in the low-risk and intermediate-risk groups, while the intermediate-risk group had a slightly higher cumulative incidence than the low-risk group. While the high-risk group's cumulative mortality rate remained significantly higher than the other two groups, the intermediate-risk group's mortality rate began to surpass that of the low-risk group, demonstrating a more pronounced divergence. This phase showed a gradual divergence in the cumulative mortality rates between the intermediate-risk and low-risk groups, with the high-risk group experiencing a continued increase in its risk of death.

[0086] By 2013, the mortality risks of the three groups had further diverged, forming a distinct mortality stratification. Clear differentiation was observed between the low-risk, medium-risk, and high-risk groups. The high-risk group had a much higher cumulative mortality rate than the medium- and low-risk groups, demonstrating a significant high-risk profile. The medium-risk group had a higher cumulative mortality rate than the low-risk group, indicating a long-term trend of increasing mortality in the medium-risk group. The low-risk group had the lowest cumulative mortality rate, indicating a better survival expectancy.

[0087] These results demonstrate the effectiveness of the obesity genetic risk model in long-term prediction, particularly its ability to identify and categorize high-risk individuals, providing a scientific basis for early intervention and risk management. Furthermore, the model not only distinguishes high-risk from low-risk individuals but also provides effective predictions for moderate-risk groups, contributing to more precise personalized medical management strategies.

[0088] It is understandable that the obesity genetic risk model provided by the embodiment of the present application combines multiple machine learning algorithms to improve prediction accuracy through a hierarchical structure. The basic learner processes the genetic feature data respectively, and the meta-classifier is responsible for integrating the prediction results and finally outputting a comprehensive score (OPRS). The quartile division method of OPRS provides a clear obesity risk classification for an individual, which helps to guide the individual in health risk management.

[0089] S600: Obtaining a model evaluation result of the obesity genetic risk model, and optimizing parameters of the obesity genetic risk model according to the model evaluation result.

[0090] In an embodiment of the present application, the method for constructing an obesity genetic risk model based on a stacking model includes, in step S600 , obtaining a model evaluation result of the obesity genetic risk model, and optimizing the parameters of the obesity genetic risk model according to the model evaluation result.

[0091] It is understandable that obtaining the evaluation results of the obesity genetic risk model is a key step in optimizing the model. It is understandable that model evaluation mainly quantifies the performance of the model by calculating common evaluation indicators such as accuracy, recall, F1 score and area under the ROC curve (AUC). These indicators can comprehensively evaluate the classification effect of the model, especially for complex problems such as obesity genetic risk that are affected by multiple factors. Based on the evaluation results, the parameters of the model can be fine-tuned to ensure that the model performs more stably on different data sets, avoid overfitting or underfitting problems, and further improve the generalization ability and prediction effect of the model.

[0092] Specifically, obtaining a model evaluation result of the obesity genetic risk model includes evaluating the model evaluation result of the obesity genetic risk model according to a model evaluation method. The model evaluation method includes calculating the accuracy, recall, F1 score, and area under the receiver operating characteristic (ROC) curve of the obesity genetic risk model.

[0093] It is understandable that, in the present application embodiment, obtaining the model evaluation result of the obesity genetic risk model includes measuring the performance of the model according to a series of evaluation indicators. Specifically, these evaluation methods include accuracy (for measuring the overall accuracy of the prediction results), recall rate (for measuring the sensitivity of the model in detecting obesity risk), F1 score (taking into account the balance of accuracy and recall rate), and the area under the ROC curve (AUC, for evaluating the discrimination ability of the model under different thresholds). Through these evaluation indicators, the performance of the obesity genetic risk model in different scenarios can be fully understood, and the model can be optimized if necessary.

[0094] In an embodiment of the present application, the meta-classifier is a logistic regression model, which generates a prediction value of obesity genetic risk by integrating the prediction results of the basic learner.

[0095] As you can understand, the meta-classifier uses a logistic regression model, which integrates the predictions of all base learners to generate the final prediction value for obesity genetic risk. The core function of the meta-classifier is to perform a weighted integration of the outputs of the base learners, compensating for potential biases in a single model and thus improving the model's overall predictive performance. Because the logistic regression model can process linear combinations of features, it is a commonly used meta-classifier in ensemble learning, helping to improve the model's interpretability and robustness.

[0096] In an embodiment of the present application, a first genetic feature set is constructed by using a random forest algorithm to obtain genetic feature importance scores, and genetic features with importance scores above a preset threshold are selected as the first genetic feature set. A second genetic feature set is constructed by using a distributed gradient boosting library model to obtain genetic feature importance scores, and genetic features with importance scores above a preset threshold are selected as the second genetic feature set.

[0097] It is understood that the first and second genetic feature sets were screened using the random forest algorithm and the distributed gradient boosting library model, respectively. The random forest algorithm calculated the importance score of each genetic feature and selected features with scores above a preset threshold to form the first genetic feature set. Similarly, the distributed gradient boosting library model also screened features based on their importance scores, with features with high scores being included in the second genetic feature set. This combined screening method ensures that key features highly correlated with obesity risk are retained in the genetic feature set.

[0098] In an embodiment of the present application, the method further includes randomly dividing the set of significant genetic features into five subsets of equal data size. The four subsets are sequentially used to train the base learner, and one subset is used to validate the obesity genetic risk model, thereby forming a cross-validation model. After completing five cross-validations, five sets of predicted values ​​are obtained, and the five sets of predicted values ​​are used to form a training set.

[0099] It is understandable that five cross validation techniques have been applied in the present application embodiment to ensure the generalization ability and robustness of the model. Specifically, the significant genetic signature set can be randomly divided into five subsets with equal data amounts. In each iteration, four subsets are used as training sets to train the basic learning device, and the remaining subset is the most verification set for verifying the performance of the model. By five cross validations, the overfitting problem of the model can be effectively prevented, while improving its performance consistency on different data sets. The five groups of predicted values ​​finally obtained will be combined into new training sets, further improving the robustness of the model.

[0100] As you can understand, in each iteration, the base learner is trained using the training set and generates predictions on the validation set. After five iterations, the five sets of predictions are aggregated to form a new training set for the base learner, which serves as the input to the meta-classifier. This cross-validation enhances the model's stability and generalization ability, ensuring consistent performance across different datasets.

[0101] In an embodiment of the present application, the data set includes an internal test cohort and an external test cohort. The method further includes inputting the internal test cohort into each base learner to obtain five sets of internal prediction values, and taking the average of the internal prediction values ​​as the internal test set. Inputting the external test cohort into each base learner to obtain five sets of external prediction values, and taking the average of the external prediction values ​​as the external test set. Outputting obesity genetic risk prediction results based on the external test set input and the distributed gradient boosting library model.

[0102] As expected, the performance of the obesity genetic risk model was also evaluated using internal and external test cohorts. Each base learner generated five sets of predictions for the internal and external test cohorts, respectively, and the average of these predictions was taken as the new internal or external test set. The average of the external test set was used as input and combined with the output of the distributed gradient boosting library model to generate the final prediction for obesity genetic risk. This approach not only validated the model's performance on internal data but also tested its generalization ability on the external test set, ensuring its robustness under different data distributions.

[0103] In the present application embodiment, the obesity genetic risk model construction method based on the stacking model has good scalability and adaptability. Specifically, as new data continues to accumulate, the training queue and test queue are updated regularly to ensure that the model is always trained and optimized based on the latest data. Through the incremental learning method, the model parameters are gradually adjusted to adapt to the changing trend of the data. In addition to single nucleotide polymorphisms (SNPs), feature fusion is performed in combination with other genetic characteristics (such as gene expression levels, metabolite data) and environmental factors (such as eating habits, exercise volume). Through the comprehensive analysis of multidimensional features, the model's prediction accuracy for obesity risk is improved. According to the differences in genetic characteristics of different populations (such as different races and age groups), the parameter settings of the stacking model are optimized. Sub-models specifically for specific populations are developed to ensure the applicability and accuracy of the model in different populations. A real-time monitoring system is developed to continuously track the health data of individuals and dynamically update their OPRS scores. Through the real-time feedback mechanism, health management strategies are adjusted in a timely manner to improve the effectiveness of intervention measures. Through the above-mentioned expansion and optimization measures, the stacking model can better adapt to different application scenarios and population needs, and continuously improve the accuracy and practicality of obesity genetic risk prediction.

[0104] The stacked model-based obesity genetic risk model construction method provided in the embodiments of the present application improves the accuracy and stability of obesity genetic risk prediction by integrating multiple machine learning algorithms and combining meta-classifier logistic regression for stacked training. Through multiple feature screening and cross-validation, the model is ensured to have good generalization ability and robustness. At the same time, the model's interpretability and transparency design enhance users' trust in the model's prediction results, providing a scientific basis for individualized health management and public health decision-making.

[0105] Figure 5 This is an electronic device 20 provided in one embodiment of the present application. Figure 5 As shown, the electronic device 20 includes at least the following parts: a processor 21 and a memory 22 .

[0106] In the embodiment of the present application, the memory 22 is used to store instructions executable by the processor 21. The processor 21 is configured to execute the instructions to implement the following Figure 1 The method for constructing an obesity genetic risk model based on a stacking model is shown.

[0107] In an embodiment of the present application, a computer-readable storage medium includes instructions, the instructions instructing a device to execute the following Figure 1 The method for constructing a genetic risk model for obesity based on a stacking model is described in [1]. For example, the instruction instructs the device to execute the following Figure 1 The method for constructing an obesity genetic risk model based on a stacking model is shown in steps S100 to S600.

[0108] The program running in the electronic device 20 involved in one embodiment of the present application can be a program that controls a central processing unit (CPU) and the like to implement the functions of the above-mentioned embodiment involved in one embodiment of the present invention (a program that causes a computer to function). Then, the information processed by these devices is temporarily stored in a random access memory (RAM) during processing, and then stored in various ROMs such as read-only memory (Flash ROM) and a hard disk drive (HDD), and is read, modified, and written by the CPU as needed.

[0109] It should be noted that a portion of the electronic device 20 of the above embodiment may also be implemented by a computer. In this case, a program for implementing the control function may be recorded on a computer-readable recording medium and implemented by reading the program recorded on the recording medium into a computer system and executing it.

[0110] It should be noted that the "computer system" mentioned here refers to the computer system built into the electronic device 20, and is a computer system comprising hardware such as an operating system and peripheral devices. Furthermore, "computer-readable recording medium" refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into the computer system.

[0111] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when transmitting programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memory within computer systems serving as servers or clients in this context. Furthermore, the aforementioned program may be a program for implementing a portion of the aforementioned functions, or a program that can achieve the aforementioned functions by combining with programs already stored in a computer system.

[0112] Furthermore, the electronic device 20 in the above-described embodiment can also be implemented as a collection of multiple devices (device group). Each device constituting the device group may have some or all of the functions or functional blocks of the electronic device 20 in the above-described embodiment. A device group only needs to have all of the functions or functional blocks of the electronic device 20.

[0113] It is understandable that the obesity genetic risk model construction method, electronic device and storage medium based on the stacking model provided in the embodiment of the present application, by integrating multiple basic learners such as random forest, distributed gradient boosting library model, adaptive boosting and logistic regression, combined with meta-classifier logistic regression to carry out stacking training, effectively merges the advantages of different algorithms, and improves the accuracy and stability of obesity genetic risk prediction. Single nucleotide polymorphism data are multiplexed by random forest and distributed gradient boosting library models, and the intersection is taken to obtain a significant genetic feature set to ensure that the features screened out are highly correlated with obesity genetic risk, further improving the feature extraction efficiency of the model. Five cross validations are introduced in the training process of the first layer of basic learners to avoid overfitting problems, and to improve the generalization ability of the model on different data sets, thereby improving the stability and reliability of the model. Through the multi-layer training of the stacking model, the prediction results of the basic learners provide richer training data for the second layer XGBoost model, and the model performance is more robust after hierarchical optimization, which enhances its ability to predict obesity genetic risk. The model is evaluated using indicators such as accuracy, recall, F1 score, and area under the ROC curve, and the model parameters are optimized based on the evaluation results to ensure that the model can maintain good performance in practical applications.

[0114] Those skilled in the art should recognize that the above embodiments are merely intended to illustrate the present application and are not intended to limit the present application. As long as they are within the spirit of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. A method for constructing an obesity genetic risk model based on a stacking model, characterized in that: The method comprises: Acquiring a data set, the data set comprising genetic characteristic data, the genetic characteristic data comprising single nucleotide polymorphisms; Applying a random forest to screen single nucleotide polymorphisms to obtain a first genetic feature set associated with the genetic risk of obesity, specifically, the random forest is used to calculate the importance score of each single nucleotide polymorphism, and single nucleotide polymorphisms with scores above a preset threshold are screened to form the first genetic feature set; Applying a distributed gradient boosting library model to screen single nucleotide polymorphisms to obtain a second genetic feature set associated with the genetic risk of obesity. Specifically, the distributed gradient boosting library model is used to obtain importance scores of single nucleotide polymorphisms, and single nucleotide polymorphisms with scores above a preset threshold are screened to form the second genetic feature set; Obtaining a significant genetic feature set based on the first genetic feature set and the second genetic feature set, specifically comprising: taking an intersection of the first genetic feature set and the second genetic feature set to obtain the significant genetic feature set, or analyzing the first genetic feature set and the second genetic feature set according to lasso regression to obtain the significant genetic feature set; Constructing an obesity genetic risk model, wherein the obesity genetic risk model is a stacking model, and the obesity genetic risk model includes a basic learner and a meta-classifier; The basic learners include: random forest, distributed gradient boosting library model, adaptive boosting and logistic regression; The algorithms in the basic learner predict the genetic risk of obesity according to the significant genetic feature set, and output the prediction results to the meta-classifier respectively, and the meta-classifier is used to output the genetic risk prediction results of obesity; Obtaining a model evaluation result of the obesity genetic risk model, and optimizing the parameters of the obesity genetic risk model according to the model evaluation result.

2. The method for constructing an obesity genetic risk model based on a stacking model according to claim 1, wherein: The meta-classifier is a logistic regression model, which generates a predicted value of obesity genetic risk by integrating the prediction results of the basic learner.

3. The method for constructing an obesity genetic risk model based on a stacking model according to claim 1, wherein: The method further comprises: Randomly dividing the significant genetic feature set into five subsets with equal data amounts; Sequentially applying the four subsets to train a basic learner, and applying one subset to validate the obesity genetic risk model to form a cross-validation; After completing five cross-validations, five sets of prediction values ​​are obtained; A training set is formed based on the five sets of predicted values.

4. The method for constructing an obesity genetic risk model based on a stacking model according to claim 1, wherein: The data set includes an internal test queue and an external test queue, and the method further includes: Inputting the internal test queue into each base learner to obtain five sets of internal prediction values, and taking the average of the internal prediction values ​​as the internal test set; Inputting the external test queue into each of the basic learners to obtain five sets of external prediction values, and taking the average of the external prediction values ​​as the external test set; The obesity genetic risk prediction result is output based on the external test set input and the distributed gradient boosting library model.

5. The method for constructing an obesity genetic risk model based on a stacking model according to claim 1, wherein: The obtaining of the model evaluation result of the obesity genetic risk model includes: Evaluating a model evaluation result of the obesity genetic risk model according to a model evaluation method; The model evaluation method includes calculating the accuracy, recall rate, F1 score and area under the ROC curve of the obesity genetic risk model.

6. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to implement the method for constructing an obesity genetic risk model based on a stacking model as described in any one of claims 1 to 5 when executing the instructions.

7. A computer-readable storage medium, characterized in that The method comprises instructions for instructing a device to execute the method for constructing an obesity genetic risk model based on a stacking model as described in any one of claims 1 to 5.

Images