Concept Drift Detection in Prediction Classifiers for Determining Gender in Metabolomics Analysis

Abstract

Currently, one of the most challenges in data analysis is connected to prediction modeling including dynamic information. Metabolomics analysis focuses on data presented dynamic information in real-time such as time-series data. Unfortunately, prediction models based on time series data are often affected by a phenomenon called concept drift. This phenomenon can reduce the accuracy of prediction models which is an unwanted effect. On the other hand, concept drift analysis can be useful in finding confounding factors. This study is divided into two parts. The first part presents the modeling of prediction classifiers based on metabolite data. The second part of this study brings concept drift detection in the created classified models. This study presented approaches to identify one of the confounding factors in human biology.