Distributed Genetic Algorithm for Feature Selection

TL;DR

This paper demonstrates that process-based parallelism significantly accelerates genetic algorithms for feature selection and improves machine learning model performance across various metrics.

Contribution

It introduces a parallel implementation of genetic algorithms for feature selection that enhances speed and model accuracy.

Findings

Speedup of 2x to 25x with parallelism

Improved ML performance metrics (F1, Accuracy, ROC-AUC)

Empirical validation of parallel GA benefits

Abstract

We empirically show that process-based Parallelism speeds up the Genetic Algorithm (GA) for Feature Selection (FS) 2x to 25x, while additionally increasing the Machine Learning (ML) model performance on metrics such as F1-score, Accuracy, and Receiver Operating Characteristic Area Under the Curve (ROC-AUC).