A FAST CORRELATION FILTER BASED GRADIENT BOOSTING CLASSIFIER FOR DISEASE DIAGNOSIS

Abstract

Disease diagnosis is the process to find the disease with specified details of a person's symptoms. Diagnosing the Disease is time consuming due to the need to analyze relevant microorganisms. Due to large growth in world’s population, Classification model receives a great deal in any domain of research and also a consistent tool for medical disease diagnosis. The domain of classification approach is used in the disease diagnosis, disease prediction, bio informatics and so on. However, an effective disease diagnosis model and the accuracy with the disease prediction were compromised. In order to obtain higher classification accuracy for heart and stroke disease diagnosis, a Fast Correlation Filter based Gradient Boosting Classifier (FCF GBC) technique is introduced. The main objective of the FCF-GBC technique is effectively performs disease diagnosis with two processing steps. Initially, Fast Correlation Filtering (FCF) algorithm is used to select the most relevant attributes (i.e. features) for disease diagnosis and filter out the irrelevant attributes in dataset. FCF uses symmetrical uncertainty to calculate the dependences of attributes and discovers the relevant attributes. After that, A Gradient Boosting Classifier is used for classifying and predicting the heart and stroke disease from the extracted attributes. Experimental evaluation is carried out using Statlog heart disease dataset and International Stroke Trial Database on the factors such as classification accuracy, classification time, error rate and true positive rate with respect to number of patients.