A Novel Approach to Predict Cardio Disease Using Naive Bayes Algorithm

Abstract

The project aims to generate a model to predict the likelihood of an individual being affected by heart
disease in the next 10 years. Many hospitals generate data about patients that can identify early
symptoms by identifying the patterns and facts present in them. This huge data makes it difficult for the
Doctors to identify patterns. As the old saying goes “Prevention is better than Cure”, early detection
and continuous management by clinicians will bring down the mortality rate. All that said, it's
impractical to observe patients daily to accurately predict and provide 24x7 consultations, since it
requires heaps of knowledge and time. During this paper we have developed and researched models to
predict heart disease through multifarious heart attributes of patients and identified machine learning
techniques just round the corner to effectively predict heart diseases. Data Science, with its vast applied
capabilities, has a key role in processing mammoth amounts of data to extract meaningful insights from
it in every walk of life, with specific mention of healthcare services.
In this paper the Naive Bayes Algorithm is used and it is built on the Framingham dataset. And this
data is divided as training of data and testing of data in ratio of 80:20. The training set is used to train
the classifier and test data is used to generate a confusion matrix to calculate accuracy. Several
techniques can be used to solve this problem. The proposed Naive Bayes algorithm has achieved an
accuracy of approximately 81.5% on the testing data, which is greater than Decision Tree accuracy.