Meta-classifier approach with ANN, SVM, rotation forest, and random forest for snow cover mapping

Abstract

This study proposes a meta-classifier approach to combine several individual classifiers for improving the accuracy of snow cover prediction. The results of the proposed approach were compared with the state-of-the-art classifiers: artificial neural network, support vector machine, rotation forest, and random forest. Past studies indicate that such approach has been rarely used for snow cover mapping. This study was conducted in the surrounding region of Gomukh, Uttarakhand, India. The base classifiers were trained on the Landsat ETM+ multispectral images. The performance of the proposed approach was evaluated based on several statistic parameters and receiver operating characteristic (ROC) curve. This study indicates that the proposed model outperformed the recent used state-of-the-art learning algorithms. © Springer Nature Singapore Pte Ltd. 2018.