Landslide Classification Using Deep Convolutional Neural Network with Synthetic Minority Oversampling Technique

Landslides are one of the world's most devastating and catastrophic natural disasters affecting human life and the economy. Many machine learning-based studies are reported on analyzing, classifying, and predicting Landslides, but there are countless avenues where these techniques must be developed to their full potential. This work proposes a deep convolutional neural network for classifying landslide data. The synthetic minority over-sampling method is employed on the dataset to address the class imbalance issue. A total of six shallow-learning algorithms and one deep-learning algorithm were used for baseline comparison. The proposed DCNN approach outperformed all the baselines chosen with an improvement of 2.1% in the f1-score. This study shows that deep learning would be better for building models capable of classifying landslides on real-world datasets.