Multi Agent based Prediction of Heart Disease: Efficient Calculation of Execution Time and Accuracy

The present work has used Twenty-Six different Machine Learning Algorithms for prediction of heart disease by using Hybrid dataset (for heart disease). Accuracy, Balance Accuracy, ROC-AUC, F1-Score and execution time (time complexity) is calculated and compared for all Twenty-Six Algorithms. Multi Agent System is developed and Multiple Agents are trained with the best fit Algorithm. It is tested that the trained Agents can predict the heart disease with almost the same accuracy. Five heart disease datasets Cleveland, Hungarian, Long beach, Stalog (heart) and Switzerland are combined to achieve better accuracy for the prediction of heart disease. Feature selection is done by ANOVA f-test at the time of preparation of data. Here the dataset is split into two segments. One is training the model and another one is testing.70-80 percent of data has been kept for training and 20-30 percent for testing. The present work has used 67% data for training and the rest 33% of data for testing. Twenty Six Machine Learning based Algorithm is implemented by Jupyter notebook. Accuracy, Balance Accuracy, ROC-AUC, F1-Score and Execution time is calculated and compared for all Twenty-Six Algorithms. RandomForestClassifier() is proved as the best accurate algorithm (93.8%) for the present dataset. The best performing algorithm is implemented using Multi Agent System and achieved accuracy of approximately 96.8%. The work compares the accuracy of twenty-six different ML algorithms applied to the heart disease dataset. The best performing algorithm, achieving 93.8% accuracy, is integrated into the Multi Agent System and achieved accuracy of approximately 96.8% in predicting heart disease. Heart disease prediction in combination with Multi Agent System and Twenty-six different Machine Learning Algorithms is an innovation in the field of Machine Learning. The work has used a large dataset combining five standard sources and implemented using Jupyter Notebook.