A cost-sensitive meta-learning classifier: SPFCNN-Miner

Abstract

Abstract Classification is a data mining technique that is used to predict the future by using available data and aims to discover hidden relations between variables and classes. Since the target data are high-dimensional, limited and class-unbalanced distribution in most real-world classification, most conventional classification methods can hardly achieve good classification results on these data. To explore an effective solution, this paper proposes the Siamese Parallel Fully-connected Neural Network (SPFCNN) as a binary classifier and uses the meta learning method to deal with the problem of class-unbalanced data distribution. Given that classified cases naturally come with costs, cost-sensitive learning is used to transform our cost-insensitive SPFCNN into cost-sensitive SPFCNN which is suitable for the classification of cost-sensitive issues. An extensive computational study is also performed on cost-insensitive and cost-sensitive versions of the proposed SPFCNN and effective results on different versions of SPFCNN which are obtained show that the performance of the proposed approach is better than that of the comparison methods.