Enhancing Machine Learning Model Accuracy with Effective Data Scaling Strategies

Abstract

The study on the impact of data scaling techniques on machine learning algorithms for predicting heart disease highlights the importance of preprocessing in enhancing model performance. Data scaling is essential when dealing with datasets that have diverse attribute ranges, as it can significantly influence the effectiveness of various machine learning models. In this investigation, eleven widely used algorithms, including K-Nearest Neighbors (KNN) and Logistic Regression, were evaluated using three scaling methods: Min-Max scaling, Z-score standardization, and MaxAbs scaling. The performance was assessed through precision, recall, andF1 score metrics across multiple experiments.The findings indicate that several algorithms performed betterwith MaxAbs scaling, particularly those sensitive to data distribution, such as KNN and Logistic Regression.This suggests that the choice of scaling technique is crucial for achieving accurate and consistent predictionsin machine learning applications related to heart disease. The results emphasize the need for careful selectionof scaling methods to optimize the performance of machine learning models in medical diagnostics.