Chronic Disease Prediction using Machine Learning

Abstract

Chronic diseases represent a major health burden worldwide. Machine learning techniques have been extensively used in the medical field to diagnose chronic diseases. Early disease diagnosis and treatment reduce the threat of having further severity of the disease, and hence, associated mortality. newlineThe main objective of this research is to propose a method that reduces the dimensionality of medical datasets by removing irrelevant and redundant features that improves the classification accuracy and, at the same time, reduces the computational time. To achieve the desired objectives, this research first propose an efficient hybrid dimensionality reduction method consisting of ReliefF and PCA method. The key aspect is the selection of an appropriate threshold for the elimination of irrelevant and redundant features from the dataset. The presented work is suitable for both text and micro-array datasets that shows remarkable results with different chronic disease datasets. newlineSecond, a novel adaptive classification system using Support Vector Machine (SVM) Classifier is proposed for the diagnosis of chronic diseases. The generalization performance of SVM classifier highly depends on the appropriate setting of its hyperparameters. The tuning of SVM parameters with Radial Basis Function (RBF) kernel is performed using the grid search method and 10 fold cross validation to obtain the best values for the considered parameters that are calculated based on Mean Absolute Error (MAE) , Classification Accuracy and Computational Time . newlineFurther, to overcome the major challenge of faster processing of medical datasets, the proposed approach uses GPUs to concurrently run different processes on machine workers. The experimental findings with benchmark datasets indicate that the proposed model implemented through parallel execution yields significant improvement in time, when compared to the conventional approaches. newlineThe benchmarked chronic disease datasets from UCI machine learning repository, Kent Ridge Biomedical Repository