Enhancing Diabetes Detection Through Data Preprocessing: A Comparative Analysis of Machine Learning Algorithms

Abstract

This study evaluates the performance of K-means, SVM, and KNN algorithms for diabetes detection using the PIMA Indian database. We apply data preprocessing techniques like PCA, LDA, and various scaling methods. Results indicate that data preparation critically impacts algorithm effectiveness. SVM, combined with PCA and MinMaxScaler, achieves the highest accuracy. KNN also benefits from PCA and MinMaxScaler. K-means shows the greatest improvement, with the WAF1-Score rising from 0.56 to 0.76 when paired with PCA and Robust Scaler.