Attribute reduction for dynamic data sets

Many real data sets in databases may vary dynamically. With such data sets, one has to run a knowledge acquisition algorithm repeatedly in order to acquire new knowledge. This is a very time-consuming process. To overcome this deficiency, several approaches have been developed to deal with dynamic databases. They mainly address knowledge updating from three aspects: the expansion of data, the increasing number of attributes and the variation of data values. This paper focuses on attribute reduction for data sets with dynamically varying data values. Information entropy is a common measure of uncertainty and has been widely used to construct attribute reduction algorithms. Based on three representative entropies, this paper develops an attribute reduction algorithm for data sets with dynamically varying data values. When a part of data in a given data set is replaced by some new data, compared with the classic reduction algorithms based on the three entropies, the developed algorithm can find a new reduct in a much shorter time. Experiments on six data sets downloaded from UCI show that the algorithm is effective and efficient.