FILTER-WRAPPER INCREMENTAL ALGORITHM FOR ATTRIBUTE REDUCTION IN INCOMPLETE DECISION TABLES WHEN OBJECT SET AND ATTRIBUTE SET CHANGE VALUE

Abstract

In the development trend of big data, decision tables are often incomplete, increasingly large in size and always changing and updating. The construction of incremental algorithms efficiency according to the filter - wrapper approach to minimize the number attribute of reduct, thereby improving the efficiency of classification and machine learning models is a very important research issue. In this paper, we propose two distance based filter-wrapper incremental algorithms: the IFWA_U_Obj algorithm in case the object set change value and the IFWA_U_Attr algorithm in case attribute set change value. Experimental results show that proposed filter - wrapper incremental algorithm decreases significantly the number of attributes in the reduct and improves classification accuracy compared to filter incremental algorithms reported.