Algorithm 2: CEDA for RMA.

Input: Training data   Output: Predictive inferences   Search for one major factor (a covariate feature-set) for the response feature-set   based on the MCE roadmap and build a geometry of response manifold by coupling response and major factor feature-sets.   Partition the entire response manifold into a collective of hypercubes as localities.   Identify a minor factor based on Shannon entropy within each locality.   Find localities containing the vector of a major factor pertaining to testing data   and predict each locality by incorporating information provided by major factor   and extra information from minor factor.