Comparative Analysis of optimized Machine Learning Algorithms for Heart Disease Severity Recognition Using Spectrophotometric analysis

The development of machine learning technology and its algorithms has led to a significant breakthrough in the medical field. The ability to diagnose and predict diseases with high accuracy has been achieved with the help of machine learning. This capability is particularly useful in creating systems that can automatically analyze medical test data without the need for a doctor’s presence. This paper focuses on designing a system for analyzing data obtained from spectrophotometric analysis to diagnose heart diseases. Different classification methods such as K Nearest Neighbor, Parzen, Bayesian, Multilayer Perceptron, and RBF have been used in this research. It also suggests unbalanced, non-normalized, and whitened data as pre-processing methods. In addition, Genetic algorithm has been used to optimize hyperparameters. Experimental reports and results show that KNN, Parzen, One-layer perceptron and Two-layer perceptron classifiers have an accuracy of 75.64%, 80.67%, 84.74%, and 71.66%, respectively. And the RBF neural network and the Bayesian method with whitened data and optimized parameters have higher accuracy than other methods and can detect the subject’s health or disease with 100% accuracy and analysis providing the expert audience with a detailed description of the methods used to diagnose heart diseases.