Uncovering Potential Attribute Relevance via MIA-Processing in Data Mining

The purpose of a classification learning algorithm is to accurately and efficiently map an input instance to an output class label, according to a set of labeled instances. In which data preprocessing, especially feature selection (FS) and continuous feature discretization (CFD), are considered as the significant issues. Since the quality of the data highly affects the result of a learning problem. Especially in medical domain, symptoms are interacted with each other; a compound symptom always could reveal more accurate diagnostic results. Therefore, a useless attribute by itself may become potentially relevant by providing hidden supportive information to other attributes. In this paper, our MIA-processing methods focus on uncovering hidden attributes relevance during FS and CFD. Our methods hence minimize the uncertainty and at the same time maximize the final classification accuracy. The empirical results demonstrate a comparison of performance of various classification algorithms on several real-life datasets from UCI repository