EFFICIENT CLASSIFICATION ON THE BASIS OF DECISION TRESS

Abstract

Machine learning has gained a great interest in last decade among the researchers. Having such a great community which provide a continuously growing list of proposed algorithms, it is rapidly finding solutions to its problem. Therefore more and more people are entering the field of machine learning to make the idea of machine learning algorithms more useful and reliable for the digital world.

In this thesis , efficient and accurate classification performances by various machine learning algorithms on the proposed heart disease prediction dataset is being discussed.The dataset cosists of 14 cloumns and 1026 rows. First 13 columns consists of Predictor variables and the last column is our target variable, which consists of Categorical values (0 and 1). Our main focus is to find the best fit model on the above proposed dataset. The machine learning algorithms that are applied and compared are KNN, SVM and DECISION TREES CLASSIFIERS. The programming tool used is python 3.7 and jupyter notebook installed in our systems. The libraries that are installed from external source to our system for plotting decision trees is graphviz and pydotplus. This thesis comprises of six chapters. Chapter 1 is our introduction part to have brief review about the used technologies in our research work process. Chapter 2 is our literature review section which basically focuses about all the past works that have been made till date on the above mentioned dataset and research ideology. Chapter 3 consists of proposed machine learning algorithms workings and their respective Methodologies in detail. (v) Chapter 4 is our results and discussion part in which we have evaluated the Performances of various machine learning algorithms used in our research work and finding out best performing Algorithm. Chapter 5 is our conclusion section in which we have concluded the best fitting Machine learning algorithm model on the basis of results and discussions in chapter 4. Chapter 6 is our refrences section. The brief review of this thesis is , to find the best fit model With highest accuracy and with minimum number of misclassification on the given heart disease Prediction dataset.