Identifying and Rectifying Rational Gaps in Fuzzy Rule Based Systems for Regression Problems

Fuzzy Rule Based Systems (FRBSs) can suffer from incomplete and sparse rule bases as a result of selecting a small number of rules from a large universe of potential rules. This may lead to rational gaps creeping into the input output mapping, where sometimes, strongly correlated inputs displaying a linear relationship with the output do not exhibit the same behaviour during inferencing. This paper proposes a technique for identifying and rectifying such gaps for FRBSs using incomplete rule bases in real-world regression problems.