Prediction of risk of heart disease for diabetic patients using data mining

Abstract

Data mining for healthcare is useful in evaluating the effectiveness of medical treatments and it is an interdisciplinary field of study that has its roots in databases statistics machine learning and data visualization Diabetic heart disease refers to the heart disease that develops in persons with diabetes The term diabetes is a chronic disease that occurs either when the pancreas does not produce enough insulin or when the body cannot use the insulin that is produced effectively Heart disease or cardiovascular disease is the class of diseases that involves the heart or blood vessels Even though many data mining classification techniques exist for the prediction of heart disease there is insufficient data for the prediction of heart diseases in a diabetic individual This thesis seeks to create both theoretical and product oriented framework which is in particular applicable to Chennai Tamilnadu India This research pertains to a prediction model which initiates Heart Disease Risk Prediction Model HDRPM using data mining classification techniques The main objective focus on this research is to find an optimal model and test the ability of classification algorithms with state of the art parties in global health care domain A number of experiments have been conducted using Weka and Rapid miner tools for comparison of the performance of predictive data mining techniques on the diabetic dataset with 1000 records using different attributes In the first experiment naïve Bayes data mining classifier technique has been applied in Weka tool which produces an optimal prediction model using minimum training set In the second experiment support vector machine data mining classifier technique has been applied in Weka tool with radial basis function kernel to diagnose vulnerability of diabetic patients to heart diseases In the third experiment in this work Rapid miner has been used as a tool and it aims to determine the most accurate technique between support vector machine and decision tree induction to predict the risk of heart disease All the above three experiments find the chances of risk in diabetic patients for heart disease using two classes high and low In the final experiment a comparative study has been carried out on the classifiers which lead to the risk of diabetic patients getting heart disease from a machine learning perspective The three chosen methods were repeatedly employed with different parameter settings to build the prediction model Some of the rules are also derived from the decision tree generated for all the models Out of the three chosen methods the decision tree provides the highest classification accuracy of ninety points seventy nine percentage The performances also have been compared using accuracy sensitivity specificity and F score Not only in overall accuracy but also in terms of precision and recall of the three classes such as high medium and low decision tree has exhibited a good performance The use of the decision tree using various split methods such as gain ratio information gain and gini index has been investigated in the thesis Decision tree model was consistent in its performance and outperformed naïve Bayes and support vector machine model So we finally fine tuned the decision tree model for optimal performance for predicting the chances of heart disease for diabetic patients Though there is availability of Cleveland Clinic Foundation heart disease dataset for the sake of determining the accuracy rate in India records of about Thousand diabetic patients have been collected from Dr V Seshiah Diabetic Research Institute in Chennai India to perform the experiments