Classification of types of forests using complementary neural networks and stackingC

Abstract

The combination between stackingC and complementary neural networks is proposed in this paper. This proposed technique is used to classify types of forests which is a multiclass classification problem. Complementary neural networks consist of two opposite neural networks trained to predict truth output and falsity output. StackingC has two levels. Complementary neural networks are applied to both levels. Uncertainty is also used to enhance the classification results. It is found that our proposed technique give better accuracy result than traditional stacking, traditional stackingC, and also the combination between stacking and complementary neural networks.