The Research Of Methods For Analyzing Gut Microbiome Data Based On Active Learning

In recent years,gut microbes have become a research hotspot in medicine and life sciences.Studies have shown that human health and even brain emotions are closely related to gut microbes.The existence or absence of gut microbial species and the relative number of various microorganisms are of great guiding significance to the diagnosis and treatment of diseases.Therefore,the introduction of machine learning methods to analyze the relationship between various diseases and gut microorganisms is very important research.There is a characteristic of the data of diseases and gut microbes:such data contains fewer tags,or it is more difficult to obtain tags.This makes traditional machine learning methods ineffective in many situations.Based on this problem,this article uses an active learning algorithm to analyze the relationship between disease and gut microbes.Based on microbiome data including species abundance and gene abundance for characterizing various diseases,multiple active learning algorithms are used to classify and predict samples,and at the same time,they are selected for examples.The results show that active learning is superior to existing methods in efficiency and accuracy.The specific contributions are as follows:Firstly,a disease prediction method based on active learning algorithm was proposed:aiming at the characteristics of less labels of diseases and intestinal microorganisms,active learning could actively select samples with large information content and send them to the trainer for iterative training.A large number of studies have shown that this algorithm performs better in the data set with fewer labels.In this paper,three kinds of active learning algorithms are mainly used to classify and predict intestinal microbial data.The results show that active learning has obvious advantages over other traditional algorithms in the efficiency and accuracy of this kind of microbial data.Secondly,a method for the selection of intestinal microorganisms and disease samples based on learning algorithm is proposed.In many patients with diseases,how to select samples with high value for disease research is of great significance for disease analysis.Some of the traditional sample selection algorithms are not effective.In this paper,the sample selection strategy of pool based active learning is applied,which mainly includes the active learning algorithm based on batch processing discrimination and representative query and the active learning algorithm based on self-paced process to select the sample of intestinal microbial data.The experimental results show that the performance of active learning in this kind of data is much better than that of other algorithms.In summary,in terms of microbiome data analysis,the active learning algorithm has many advantages over other machine learning algorithms.Active learning does not require all samples to have labels,so it requires a small number of label samples.The active learning algorithm can train the data iteratively,and actively adjust the learner after each iteration to achieve a more accurate learning effect.It is an effective model to analyzing microbiome data.