An Ensemble Classification Method for High-Dimensional Data Using Neighborhood Rough Set

Algorithm 1

Feature selection.

	Input: GEDT = {S, Pi∪D, V, f}
	δ //the list of intersection neighborhood threshold
	Output: RED // a set of features which is a reduction of unit Pi
	Step 1: For each sample skS, calculate the intersection neighborhood based on different threshold δ
	Step 2: Divide the sample set S based on the class label D = {d} to obtain the equivalence classes which are represented as S/IND(D) (where samples with the same class label contained in one equivalence class).
	Step 3: Calculate the positive region defined on intersection neighborhood based on all of the features in Pi.
	Step 4: Start with RED = Pi.
	Step 5: As Step 1 for sample skS, calculate the intersection neighborhood . Then as Step 3 calculate the positive region based on RED-{pij}, ;
	If , then make RED = RED−{pij}.
	Step 6: Repeat Step 5 until all the features pij in the subset Pi are validated, then use final RED as a reduction of Pi, marked as  = RED.