PCA based Machine Learning techniques for Heart Disease Diagnosis System

Prominent and preliminary information mining has spread to a great extent globally with innovative real-time applications. It rapidly addresses the challenges in marketing, e-business, retail, education, health sectors, etc., in different areas and determination. Medical services are among those that need quick recommendation services and diagnosis through technological upgrades. In any situation, effective investigation apparatus is limited in identifying abandoned patterns in data. With the application of machine learning, this work aims to provide a model with better performance measured based on various classification metrics like precision and accuracy.Method: The proposed work applies data preprocessing techniques like label encoding, standardization scaling and principal component analysis on the heart diagnosis dataset. Further, the model is built using various classification algorithms like SVM, KNN, Random Forest, Logistic Regression and decision tree and evaluated on unseen observations.Results: The alignment of predictive machine learning techniques within the same dataset has been examined to detect heart disease based on two binary class labels: yes and no. The observed results show better performances for Decision Tree on unscaled data and PCA (both 80%), Logistic regression with PCA (86%), Random Forest on Scaled and PCA (both 88%), KNN and SVM on PCA with scaled data (86% & 87%). The Random Forest model surpasses all other models with more accuracy.