A predictive model for heart disease diagnosis based on multinomial logistic regression

Abstract

Many computational approaches are used to assist the analysis of influencing factors, as well as for the need for prediction and even classification of certain types of disease. In the case of disease classification, the data used are often categorical data, both for dependent variables and for independent variables, which are the results of conversion from numeric data. In other words, the data used are already unnatural. Conversion processes often do not have standard rules, thus affecting the accuracy of the classification results. This research was conducted to form a predictive model for heart disease diagnosis based on the natural data from the patients' medical records, using the multinomial logistic regression approach. The medical record data were taken based on the patients’ electrocardiogram information whose data had been cleansed first. Other models were also tested to see the accuracy of the heart disease diagnosis against the same data. The results showed that multinomial logistic regression had the highest level of accuracy compared to other computational techniques, amounting to 75.60%. The highest level of accuracy is obtained by involving all variables (based on the results of the first experiment). This research also produced seven regression equations to predict the heart disease diagnosis based on the patients’ electrocardiogram data.