Rainfall prediction with TLBO optimized ANN
Rainfall prediction is very crucial for India as its economy is based on mainly agriculture. The parameters that are required to predict the rainfall are very complex in nature and also contain lots of uncertainties. Although various approaches have been earlier suggested for prediction, the soft computing is found to be very effective in developing models which emulates human being and derives expertise like human being to adapt to the situations and learn from the experiences. In this study, rainfall prediction for Andhra Pradesh (AP) state is carried out with Artificial Neural Network (ANN). A new heuristic approach Teaching Learning Based optimization (TLBO) is used to train the weights of the ANN developed for rainfall prediction. A comparison is done with classical back Propagation learning approach and mTLBO (a variant of classical TLBO). The data of monthly rainfall (mm) in Coastal Andhra is collected from Indian Institute of Tropical Meteorology (IITM), Pune, India. The data set consists of 1692 monthly observations during years 1871 to 2011. The simulated results reveal the effectiveness of ANN-mTLBO over ANN-BP on investigated datasets. The findings of our work will be very useful in assessing the possible drought situations in AP from the rainfall predictions.