ENRICHING FLOOD AND LANDSLIDE PREDICTION THROUGH DEEP LEARNING MECHANISM

Abstract

Despite their widespread distribution, regular occurrence, and numerous, geographical, and devastating qualities, landslide catastrophes have inflicted immeasurable damages to the country's economy, lives, and properties. India has intrinsic and environmental causes for landslide incidence due to its tropical humid long rainy season and unique geographical position. The rainfall-induced landslide's realization is of tremendous importance. Because the fundamental circumstances for both calamities are identical, landslides frequently precede floods or vice versa. Because of its crucial significance in decreasing economic and lives losses, flood prediction is among the most challenging, complicated, and significant challenges in engineering. Forecast accuracy has improved in recent years has resulted of advances in data collecting via satellite observations, as well as advances in technology and computational methods for uncertainty analysis and interaction. Therefore an approach for landslide and flood prediction is the need of the hour. For this purpose the proposed work is incorporates the LSTM neural network which is basically a time based model to predict the flood and landslide. This process is catalyzed by the use of K- nearest neighbor classification model which is ensemble with LSTM to produce good results of root mean square error.