Prediction Landslide Location Area Using ANN (Case study in Bangli Regency, Bali Indonesia)

Abstract

Landslides are significant geo-hazards heavily impacting many regions of the world regarding human lives and economic losses. The large magnitude of natural forces involved in landslides makes actions of mitigation or prevention unfeasible, with exceptions for minor occurrences or under special conditions. Many old methods have been applied in landslide management and/or prediction, such as overlays or weighting methods. The newest/advanced methods are still being developed and one of the newest methods is Artificial Neural Network (ANN). ANN are biologically inspired computer programs designed to simulate how the human brain processes information. Many types of ANN exist; the most famous one is Multilayer Perceptron (MLP) Neural Network Algorithm with FeedForward model. MLP consists of three parts: the input layers as neurons representing the value of data; the hidden layer, which demonstrates the network training process; and the output layer, which provides the prediction of the landslide areas. In this research, the input layer consists of landslide location characteristics, such as the rainfall intensity, land cover, slope, geological types, and rate displacement of landslides. As a case study, Bangli Regency was selected. In 2017 there was a landslide disaster in the Kintamani District, Bangli Regency, which resulted in dozens of people missing or dead, and several houses destroyed. In this study different numbers of neurons were used in the hidden layer (15, 50, 100, and 150 neurons). The best performance is obtained at 150 neurons, with 0.9677 (96,77%) for the test set.