Redefined Classification of Hepatitis Variants through Arima, Signal Processing and Machine Learning

This study introduces a novel feature selection method using the Auto Regressive Moving Average model to classify four hepatitis virus types: Hepatitis B, C, D, and E. Firstly, DNA sequences are transformed from characters into numerical representations using Electron-Ion Interaction Potential coding.Then discrete sine transform is applied to extract features from the sequences.Our proposed feature selection technique, using the inverse integrated moving average (ARIMA) as a statistical filter, refines the features for optimal classification.We train two machine learning models Light Gradient Boosting Machine and Random Forest on the selected features.We compare the results of our approach with another 5 window-based finite impulse response filters (FIR) as they extract significant features by reducing noise and emphasizing key patterns.Our findings indicate that ARIMA, with an order of (1,1,5), achieves 98% accuracy, while the Bartlett window-based filter yields 93.25% accuracy.This approach highlights the potential of ARIMA for effective feature selection in the classification of hepatitis viruses using machine learning.