A Comparative Analysis Of Feature Selection Techniques Using Automated Machine Learning Tools

Abstract

AutoML (Automated Machine Learning) is the field that seeks to automate the process of developing machine learning models. AutoML is created to boost productivity and efficiency by automating as much of the process that occurs when machine learning is applied, which streamlines the workflow from data preprocessing to model deployment, especially as it is considered important for the feature selection process. In this study, we use two popular AutoML frameworks, TPOT and KNIME, to compare numerous feature selection methods. Feature selection is a crucial step in the machine learning pipeline, as it involves identifying the most relevant features that improve models' ability. Effective feature selection can improve model accuracy, reduce overfitting, and enhance interpretability by focusing the key attributes. In this study, we used the autism spectrum disorder (ASD) dataset which is collected from multiple rehabilitation centres in Pakistan, our goal is to determine which features offer the best model for the diagnosis of autism spectrum disorder (ASD). TPOT and KNIME both demonstrated their capability in identifying ASD, achieving impressive accuracy rates of 85.23% and 83.89%, respectively. The evaluation metrics precision, recall, and F1-Score, among others— verified the models' reliability as well. The proposed frameworks and their feature selection methods enhanced the overall approach of the model in addition to identifying important features that have a strong impact on the model. Using these AutoML frameworks not only optimised the feature selection process but also greatly reduced the amount of time required for diagnosis. This study demonstrates how AutoML approaches, and feature selection techniques can be used to improve model efficiency, which will help with early detection and improve outcomes for children with ASD and their families.