Open AccessPredictive Analytics on Rainfall using Long Short Term Memory for Identification of Drought
Author(s) -
Vignesh Karthikeyan*,
S. Poornima,
M. Pushpalatha
Publication year - 2019
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.b2485.098319
Subject(s) - economic shortage , term (time) , identification (biology) , scope (computer science) , precipitation , artificial neural network , limit (mathematics) , econometrics , environmental science , computer science , climatology , geography , machine learning , meteorology , mathematics , geology , ecology , mathematical analysis , linguistics , philosophy , physics , quantum mechanics , government (linguistics) , biology , programming language
A drought is duration of below-average precipitation in a certain region, resulting in prolonged shortages in the supply of water. It occurs naturally and has perilous impacts on the society. Observation of patterns of droughts in the past and using it to predict the ones likely to occur in the future can be very helpful. Preparations can be made to try and limit their effects on the society. Drought is however random and dependent on drought variables that possess a non-linear nature. With development in neural networks in the past years, it has shown good scope for time-series prediction with non-linear models. This research approaches the drought prediction problem with the use of Recurrent Neural Networks. The proposed model makes use of past years rainfall values to predict the risk of shortage of rainfall in the given region. The model is expected to show better performance over the existing traditional methods.