Classification of categorical and numerical data on selected subset of features

Many Data Mining techniques use the whole features space in the classification process. This feature space might contain irrelevant, or redundant, features that could reduce the accuracy of classification. This paper presents an approach to select a subset of features that are most relevant to the classification application. We use a wrapper approach to search for relevant subset of features, which will be used in the classification of two datasets: categorical teachers’ dataset and numerical image dataset. Naïve Bayesian algorithm and KNearest Neighbor algorithm are used to classify and estimate the accuracy of the categorical data and numerical data, respectively. The experimental results for both categorical and numerical datasets indicate that classification accuracy is improved by removing the irrelevant features and using only the relevant subset of the feature space.