Ensemble Classification Based on Feature Selection for Environmental Sound Recognition

Algorithm 1

The proposed MmEnFs algorithm.

Algorithm: MmEnFs (multi-models ensemble feature selection)

Input: FSM = = 1, 2, …; p denotes the kinds of feature selection methods.

£:learning algorithm method

k:the ratio of selected features

TrainD = = 1, 2, …; m denotes the number of sample of traindata.

F = , f₂,…; n denotes the numbers of features.

Output: H(x):Ensemble model

for t = 1, 2,…,p do

Step 1: SubF = Feature selector (TrainD, FS, k)

Step 2: Sub = map (TrainD, SubF)

Step 3: (x) = Train (Sub, £)

end for