Comparison and Evaluation of Classification Techniques in Medical domain: A Case Study Using Heart Disease Prediction and Analysis

Abstract

Heart disease affects millions of lives every year and stands as one of the most common and life-threatening health issues faced by people around the world. Detecting it early is essential to prevent associated outcomes, and machine learning is playing a key role in making this possible. Machine learning has transformed many industries by enabling systems to analyze data, recognize patterns, and make informed predictions. In healthcare, these capabilities are being used to develop models that help diagnose diseases such as heart disease.This study focuses on comparing and evaluating five ML algorithms: RegressionByDiscretization, M5P, lazy.KStar, DecisionTable, and M5Rules for the classification of heart disease using the Weka platform. Instead of emphasizing the prediction of heart failure, the focus is on analyzing and comparing how well these models perform in processing patient data and identifying trends. The experiments utilized the UCI Heart Disease dataset comprising 918 patient records with clinical features including age, chest pain type, cholesterol levels, resting ECG, and maximum heart rate. Each model was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Relative Absolute Error (RAE), Relative Root Squared Error (RRSE), and Correlation Coefficient. Among all the evaluated models, the \textit{Regression by Discretization} approach exhibited the most accurate predictions, achieving the lowest MAE of 0.0999 and RMSE of 0.1016, with RAE and RRSE values of 20.69\% and 20.95\% respectively, and an exceptionally high correlation of 0.9988. By evaluating multiple techniques, this research highlights the strengths and limitations of each approach, offering insights into which models may be more effective for similar healthcare applications. The findings underline the importance of choosing the right model and suggest that even greater accuracy may be achieved by exploring more advanced techniques in the future.