Title: Local area rainfall prediction using hybrid approach

Authors: Bikash Kanti Sarkar

Addresses: Department of Information Technology, Birla Institute of Technology, Mesra, Ranchi – 835 215, India

Abstract: Rainfall prediction is important in many aspects of our economy and general livelihood by preventing any serious natural disasters. Numerous methodologies have been introduced but most of them cannot provide transparency of predicted outcomes. Addressing this problem, a hybrid model combining C4.5 classifier and genetic algorithm (GA), is proposed in this study in order to forecast rainfall of Ranchi City in India. More specifically, C4.5 classifier is run on the past metrological data of local area to extract rule set (knowledge). The generated rule set is then refined by the proposed GA for accurate prediction of rainfall. The model is tested on the test dataset as well as the upcoming day to day. The results show that the integrated model can forecast satisfactorily rainy days and non-rainy days (along with amount of rainfall).

Keywords: local rainfall; forecasting accuracy; C4.5 classifier; genetic algorithms; GAs; rainfall prediction; rainfall amounts; rain.

DOI: 10.1504/IJICA.2013.062988

International Journal of Innovative Computing and Applications, 2013 Vol.5 No.4, pp.213 - 227

Received: 25 Sep 2012
Accepted: 08 Jul 2013
Published online: 31 Jul 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article