A Feature Reduction Method by Grey Theory and Rough Set

In varied assessment work, feature reduction is a key cycle which benefits for reducing redundancy. It will improve the assessment efficiency, and reduce much burdensome statistical work. Grey theory and variable rough set theory are introduced to propose a feature reduction method. For the empty values of assessment features in past years, the grey theory based predicting method for imperfection information is proposed to dispose empty values. For each continuous feature, based on approximation precision, a continuous feature dispersing method for chronological series is proposed to obtain varied dispersed intervals. Based on the differentiating matrix, a variable precision rough set based feature reduction method is designed, and the antinoise capability is improved. The obtained assessment result by feature reduction is almost the same with that by all features, and the strong error tolerance capability is validated.