When diagnosing sudden mechanical failure, in order to make the result of classification more accurate, in this article we describe an affinity propagation clustering algorithm for feature selection of sudden machinery failure diagnosis. General methods of feature selection select features by reducing dimension of the features, at the same time changing the data in the feature space, which would result in incorrect answer to the diagnosis. While affinity propagation method is based on measuring similarity between features whereby redundancy therein is removed, and selecting the exemplar subset of features, while doesn't change the data in the feature space. After testing on clustering and taking the result of PCA and affinity propagation clustering as input of a same SVM classifier, we get the conclusion that the latter has lower error than the former.
A Quantitative Site-Specific Classification Approach Based on Affinity Propagation Clustering
