SCALING-UP REDUNDANT FEATURES USING FUZZY-ROUGH DATA ANALYSIS TECHNIQUES.
Feature Selection is an important task in data analysis especially in high dimensional data like health care data, genome data and financial data. A decision can be taken from the available information which should be certain and precise. In that case if the data is imprecise or noisy data, we cannot perform the optimal selection and the decision will not be accurate. For handling such types of inconsistent information, the theory proposed by Pawlak in 1982 performs very well named as Rough set or Indefinable set since it does not require additional information other than the available knowledge base. It is expressed by its set approximations known as lower and upper approximation and these approximations are constructed through equivalence relations between the variables (feature). The feature is said to be consistent if the removal of will affect the quality of data. Mathematically speaking the set contains these types of variables is called Reduct. An information system contains number of reduct set. Identification of stable reducts are essential which will enable the data scientists to bring optimal decision making. These type of reducts are known as Dynamic reducts. This paper proposes the method of finding Dynamic reducts using Fuzzy-Rough set model and it is being implemented for Heart Disease Data set.
Keywords: Information System, Set approximations, Fuzzy-Rough Set.
AMS Subject Classification: 03E20, 03E72, 03E75, 54C05, 54D05