Application of the Bees Algorithm to the Selection Features for Manufacturing Data
Data with a large number of features tend to be deficient in accuracy and precision. Some of the features may contain irrelevant information caused by data redundancy or by noise. A wrapper feature selection method using the Bees Algorithm and Multilayer Perceptron (MLP) networks is described in this paper. The Bees Algorithm is employed to select an optimal set of features for a particular pattern classification task. Each bee represents a possible set of features. The MLP classification error is computed for a data set with those features. This information is supplied to the Bees Algorithm to enable it to select the combination of features producing the lowest classification error. The proposed method has been tested on data collected in semiconductor manufacturing. The results presented in the paper clearly demonstrate the effectiveness of the method.
