A Novel Feature Selection Based Classification Algorithm for Real-Time Medical Disease Prediction

Abstract

In the current medical databases, feature extraction and disease prediction are the essential requirements to Chronic Obstructive Pulmonary Disease (COPD) and Alzheimer's diseases. Most of the medical databases have heterogeneous features with different levels of severity patterns. Feature extraction and classification of high risk patterns may have potential benefits for decision making. In the medical applications, data classification algorithms are used to detect the disease severity that can help in early prediction of new type of disease patterns. Also, machine learning algorithms are more accurate, high true positive rate and reliable for heterogeneous features. Traditional classification models such as Naive Bayes, SVM, Feed forward neural networks, Regression models, etc are used to classify the homogeneous disease datasets with limited feature space. As the size of the Alzheimer's disease patterns and its categories are increasing, traditional data classification models are failed to process the disease patterns due to inconsistent, class imbalance, and sparsity issues, which may affect the disease prediction rate and error rate. Therefore, an efficient classification model for predicting the severity level of the heterogeneous feature types is essential with high true positivity and low error rate. In this paper, a novel feature selection based classification model is proposed to improve the disease classification rate and testing the new type of disease patterns for real-time patient disease prediction. In the proposed model, a novel probabilistic based feature selection measure for classification algorithm is designed and implemented for real-time patient disease prediction using the training datasets. Experimental results show that the proposed feature selection based classification algorithm is better than the traditional algorithms in terms of true positive rate, error rate and F-measure are concerned.