Informatics Publishing Limited

Kodali Lohita ¹, Adusumilli Amitha Sree ¹, Doreti Poojitha ¹, T. Renuga Devi ¹, A. Umamakeswari ¹

Affiliations
1 School of Computing, SASTRA University, Thanjavur – 613401, Tamil Nadu, IN

Abstract

Objective: The main objective of the work is to compare the heart disease prediction accuracy of different data mining classification technique and to find the best technique with minimum incorrectly classified instances. Different classification techniques are used to predict heart disease based on the factors that cause these diseases which include family history, age, obesity and some other factors. Method: This work is carried out in three phases. The First Phase is preprocessing of data set. The attributes like trestbps, cholesterol, tpeakbps and age are normalized and missing values are handled appropriately. The second phase is feature selection. The greedy hill climbing best first attribute evaluator is used to identify the subset of attributes based on its individual prediction ability. The third phase is comparison of prediction accuracy of different techniques in literature. Findings: The work has been evaluated using the performance metrics like accuracy, specificity, sensitivity, confusion matrix to prove the efficiency of different techniques. It was concluded that the Bagging algorithm achieved highest accuracy compared with other algorithms.

Keywords

Classification, Data Mining, Heart Disease Prediction

Full Text