Research On Disease Prediction Methods Based On Human Intestinal Microbiome Dat

The disease prediction based on human gut microbiome data has the advantages of indirect detection,health warning,and data quantification.But the predictions of multilabel and multiple disease are ineffective due to the diversity of diseases and the interactions between diseases.Thus,the specific algorithms of disease prediction need to be explored urgently.To achieve the predictions of multi-label and multiple diseases,first,predicting healthy or diseased state by the strategy of binary classification.This paper divides the data sets of multi-label and multiple diseases both into healthy and diseased categories.After analyzing the advantages of models,this paper discovers the random forest model with the embedded host variable features has the obvious advantages of classification.After ranking the importance of human gut microbiome features,this paper finds the better fit between the human gut microbiome features and the prediction of healthy or diseased state by traversing the ratios of features.Thus,improving the effect of disease prediction.Besides,this paper analyzes the trend of prediction results from the perspective of feature importance.After achieving the prediction of healthy or diseased state,this paper predicts the specific diseases by the strategy of multi-label and multiple classifications.The specific key features of each disease can reduce the overlapping effects of multiple diseases on human gut microbiome.For multi-label diseases,this paper trains five models for each disease according to the variable embedded host variables,and the predicted result is determined by the voting strategy.Besides,the association rule between diseases is incorporated into the model,thus realizing the multi-label disease prediction.For multiple diseases,this paper trains model for each disease using the host variables and the specific key features to design a method of model selection,and the predicted result is determined by the predicted probability,thus realizing the multiple disease prediction.After achieving the predictions of multi-label and multiple diseases,this paper analyses the relationship between probiotics and human health.The data set is divided into healthy and diseased groups.First,this paper analyses the significant correlations between probiotics and the species of key features by chi-square test.Then,this paper analyses the significant positive and negative linear correlations between probiotics and the species of key features by linear regression analysis,thus getting the co-occurrence network between probiotics and the species of key features.Last,for healthy and diseased groups,this paper gets the probiotics that can improve human health significantly by analysing the relationship between the species of key features and human health.Thus,this paper proposes the solutions for improving human health.