Handling Incomplete Categorical Data for Supervised Learning

Classification is an important research topic in knowledge discovery. Most of the researches on classification concern that a complete dataset is given as a training dataset and the test data contain all values of attributes without missing. Unfortunately, incomplete data usually exist in real-world applications. In this paper, we propose new handling schemes of learning classification models from incomplete categorical data. Three methods based on rough set theory are developed and discussed for handling incomplete training data. The experiments were made and the results were compared with previous methods making use of a few famous classification models to evaluate the performance of the proposed handling schemes.