Prediction of Earthquake Induced Landslide Using Deep Learning Models

Earthquakes are one of the leading factors of a landslide. Over the past ten years, damages caused by earthquakes in human settlements are found to be increasing. Recently, landslide prediction using Radial Basis Function of Support Vector Machine with an accuracy of 91.2% is reported. With landslide prediction probability, there are two possibilities: a landslide occurrence or non-occurrence. In either cases, the prediction could be correct or false. In the first case, false prediction could result in loss of human life and property, while correct prediction is valuable. But in the latter case, false prediction could lead to human stress & strain, expensive disaster prevention measures and so on, while correct prediction is appreciable. Hence, to minimize losses it is essential to predict the landslide accurately. With this intention, prediction algorithms with high accuracy are developed. Landslide prediction using machine learning techniques such as Naive Bayes, Logistic Regression, Support Vector Method, and Random-forest are proposed. The accuracy reported by these techniques is low and so prediction using deep learning methods such as Convolutional Neural Network and Recurrent Neural Network is attempted. Performance measures of these methods are evaluated using the three pillars of binary classification namely accuracy, precision and recall. The results of the deep learning models outperform the machine learning models in order namely Logistic-Regression, Support-Vector-Machine, and Naive-Bayes classifier.