Diagnosis of Heart Disease Based on Electronic Health Record Using Extreme Learning Machine

Abstract

Nowadays, due to the significant growth in the use of medical and disease information recording systems, as well as people's health records, we are faced with a large amount of information and data that can be explored for purposes such as diagnosis, recognition, treatment, and prevention. It is considered an important and vital thing among diseases. Data mining and machine learning are powerful tools for refining and analyzing all kinds of medical data. Cardiovascular diseases are one of the main causes of death in the world and in Iran, and the best way to prevent them is to diagnose them in time and prevent that’s of. In the diagnosis of heart disease, factors such as blood pressure, age, gender, cholesterol levels, blood sugar, physical activity, etc. are usually taken into consideration and finally the risk of heart disease is determined. In this research, extreme learning machine (ELM) algorithm was used to predict heart disease. Then, the extreme learning machine algorithm was compared with the multilayer perceptron neural network with 1 and 100 hundred iterations, based on evaluation metrics such as accuracy, precision, and recall. Also, the computing time, which is an indicator of the amount of computing resources, was investigated. To investigate, the Kaggle data set was used, which has 70,000 samples and 11 features for predicting heart disease. after pre-processing the data, 80% of them were randomly used to train the model. Investigating and comparing the prediction results of extreme learning machine model and multi-layer perceptron neural network indicate that extreme learning machine records a better performance in predicting heart disease in the data set of this research. According to the findings of this research, extreme learning machine algorithm provides 72.07% accuracy, 70.65% recall and 72.76% precision.