AbstractsComputer Science

Data reduction is an important contribution of rough set theory in data analysis, data mining and machine learning. Most recent researches have focused on attribute reductions. However, another main part of reduction, value reduction, has barely been paid attention to for a long period of time. Value reduction can induce decision rules with logical equivalent minimal length.Some research ideas such as high frequency value reduction [15] and association rules [17] have been applied to value reduction to gain better generating performance. This research introduced a new value reduction algorithm combining rough set theory with association rules to generate decision rules from examples in format of Multiset Decision Tables (MDT)[8]. Testing with UCI machine learning database and comparing this algorithm to high frequency value reduction algorithm[15], this research indicated the fitness of the new algorithm to process large data and validation to improve efficiency.