Anisa Meidina, Zaenal Abidin

Diagnosis of Heart Disease Using Optimized Naïve Bayes Algorithm with Particle Swarm Optimization and Gain Ratio

Introduction

Diagnosis of heart disease using optimized naïve bayes algorithm with particle swarm optimization and gain ratio. Diagnose heart disease with an optimized Naïve Bayes algorithm using Particle Swarm Optimization (PSO) and Gain Ratio for feature selection, boosting accuracy to 93.44% on the Cleveland dataset.

Abstract

Purpose: This study aims to apply feature selection particle swarm optimization (PSO) and gain ratio to the naïve Bayes algorithm and gauging the level of accuracy before and after applying PSO feature selection and gain ratio to the naïve Bayes algorithm in the diagnosis of heart disease.Methods/Study design/approach: Data collection is done by using taking the Cleveland dataset obtained from the UCI machine learning repository. The data used in this study were 303 samples. The data is processed using the preprocessing stage. The naïve Bayes algorithm is used for a classifier, while PSO and gain ratio for feature selection.Result/Findings: The results of the study revealed that the classification accuracy of the naïve Bayes algorithm without the application of feature selection in the Cleveland dataset is 86.88%, while the results of the classification accuracy of the naïve Bayes algorithm after applying PSO and gain ratio in the Cleveland dataset is 93.44%. Application of PSO and gain ratio as feature selection algorithms can improve classification accuracy by 6.56%.Novelty/Originality/Value: This study combines the PSO feature selection and gain ratio on the naïve Bayes algorithm using the Cleveland dataset. The research model that was carried out was enriched by carrying out the preprocessing stages, namely data cleaning, changing the number of class labels, data normalization, and data discretization. This study shows that using a combination of the PSO feature selection algorithm and the gain ratio gives better accuracy to the naïve Bayes algorithm in diagnosing heart disease.

Review

This paper presents a study focused on enhancing the diagnostic accuracy of heart disease using an optimized Naïve Bayes (NB) algorithm. The core of the proposed method lies in integrating Particle Swarm Optimization (PSO) and Gain Ratio as feature selection techniques prior to applying the NB classifier. Using the well-established Cleveland dataset from the UCI machine learning repository, the authors demonstrate a notable improvement in classification accuracy. Specifically, the study reports an increase from 86.88% (without feature selection) to 93.44% (with PSO and Gain Ratio feature selection), indicating a 6.56% enhancement in performance. This research aims to contribute to more accurate and reliable automated diagnostic tools for cardiovascular conditions. The study's strengths include its clear objective of improving a widely used classification algorithm for a critical real-world application. The explicit mention of a comprehensive preprocessing stage, encompassing data cleaning, class label adjustments, normalization, and discretization, is commendable and contributes to the robustness of the data handling. However, several aspects could benefit from further clarification or comparison. While the 6.56% improvement is significant, the abstract lacks context regarding how the achieved 93.44% accuracy compares to other state-of-the-art machine learning models applied to the same Cleveland dataset. Furthermore, the abstract does not specify the cross-validation strategy employed, which is crucial for ensuring the reliability and generalizability of the reported accuracy on a dataset of 303 samples. Clarification on how PSO and Gain Ratio are specifically combined for feature selection, beyond their separate application, would also strengthen the methodological description. Overall, this paper offers a valuable contribution by demonstrating that a combined PSO and Gain Ratio feature selection approach can effectively improve the diagnostic accuracy of the Naïve Bayes algorithm for heart disease. The findings support the utility of intelligent feature engineering in enhancing machine learning model performance in medical diagnosis. For future work, it would be beneficial to conduct comparative analyses with a wider range of advanced feature selection and classification algorithms on diverse datasets, as well as to investigate the interpretability of the features selected by the proposed method. Despite the minor points for elaboration, the study presents a solid methodology and promising results, making it a relevant piece of research in the field of medical data analysis.

Full Text

Comments

You need to be logged in to post a comment.