A cloud based system for prediction of heart disease using machine learning algorithms

Heart disease is a major concern for every individual, especially millennials and the current generation who are seeing its forms at an early age because of improper nutrition and exercise. With the advent of new drugs and techniques to curb its effects, it is crucial to detect the possibility of a heart related disease in the earliest stage. This can be done with data mining to process clinical cardiac parameters measured from a patient. The health sector can take maximum advantage of this, leveraging algorithms to pool in patient data like age, medical history, heart rate, stress, and cholesterol levels while providing probability models for prediction of an illness. Heart disease is an umbrella term used for a variety of ailments which need to be defined meticulously to develop a computational model for analysis. This work presents a web-based application to predict the likelihood of a patient to have a heart disease. The raw data is fed to a decision tree that outputs primary clusters, which in turn are pipelined into a neural network to factor in possibilities of symptoms to calculate the probability of heart disease. The work aims to reduce the errors due to human negligence by factoring in the most detailed symptoms starting from the patient’s medical history to most recent test results for highest accuracy. The work aims to give better accuracy results for the prediction of Heart Disease. The algorithms that are used for analysis are tested on two datasets - one is the Cleveland Cardiac Center dataset and the other is the Switzerland dataset. Both the datasets give different accuracy results. The accuracy results for the Cleveland dataset are Decision Tree Algorithm: 89%, Multilayer Perceptron Algorithm: 89%, Stacking Classifier: 93%. The accuracy results for the Switzerland dataset is Decision Tree Algorithm: 89%, Multilayer Perceptron: 94% and Stacking Algorithm: 94%