Single-Layered Complex-Valued Neural Network with SMOTE for Imbalanced Data Classification
Kentarou Matsuda, Kazuyuki Murase · 2016
One of the most popular problems in the field of classification is how to treat imbalanced data. Improper treatment causes lowering of the learning accuracy and sensitivity of the classifier. In this research, we use the single-layered complexvalued neural network (CVNN) to classify imbalanced dataset. And also, in order to overcome the imbalanced data problem, we use the well-known over-sampling algorithm called the synthetic minority over-sampling technique (SMOTE) for the CVNN. The SMOTE is the technique which over-samples the minority class data. We use the five imbalanced datasets from the UCI depository and compared with real-valued neural networks (RVNN) to verify the efficiency of the CVNN with SMOTE for imbalanced data classification problems. As a result, the CVNN with SMOTE achieved better sensitivity and accuracy with most of the tested datasets than the counterpart.