An Incomplete Data Filling Approach Based on a New Valued Tolerance Relation

Abstract

In real life there are many incomplete information system, However, the traditional rough set theory is not suitable for incomplete information system. A lot of extension of the rough sets theory have been proposed based on this. In these theories, the handling of null value or missing values is the key problem. In this paper a new valued tolerance and a concept of Tolerance Degree Vector are put forward at first; moreover a new incomplete data filling approach which based on new valued tolerance and tolerance degree vector(for short "NVT-TDV") is proposed. Subsequently, two series of experiments have been carried out, one was compared the classification accuracy with other extended rough set model, the other was the data filling accuracy test based on Waikato Environment for Knowledge Analysis platform. The experimental results show that it can be adopted as a pre-processing method in data mining.