Enhancing disease prediction on imbalanced metagenomic dataset by cost-sensitive

Abstract

Imbalanced datasets usually appear popularly to many real-world applications and studies. For metagenomic data, we also face the same issue where the number of patients is greater than the number of healthy individuals or vice versa. In this study, we propose a method to handle the imbalanced datasets issues by Cost-sensitive approach. The proposed method is evaluated on an imbalanced metagenomic dataset related to Inflammatory bowel disease to do prediction tasks. Our method reaches a noteworthy improvement on prediction performance with deep learning algorithms including a MultiLayer Perceptron and a Convolutional Neural Neural Network with the proposed cost-sensitive for Metagenome-based Disease Prediction tasks.