A Hybrid Feature Reduction Approach for Medical Decision Support System

Feature reduction is essential at the preprocessing stage of designing any reliable and fast disease diagnosis model. Addressing limitations like specificity, information loss, operating NP problem in polynomial time, this paper introduces a two-step hybrid feature selection approach to identify subset most relevant contributing features each medical dataset for constructing diagnostic The concept gain used Step I select informative features, whereas correlation coefficient-based employed II retain possessing much dependency with class attribute but less among non-class attributes. In particular, both approaches are sequentially fused approximately optimal order construct better classification model terms performance time. threshold criteria decided choose appropriate from datasets. effectiveness proposed assessed using six individual competent learners one ensemble learner over seventeen datasets smaller larger dimensions. empirical results indicate that improves after selection, reducing considerable amount irrelevant redundant data.