DEVELOPMENT OF FEED FORWARD BACK PROPAGATION NEURAL NETWORK WITH BEST FITTING MODELS TO PREDICT SEASONAL RICE PRODUCTION IN TAMILNADU

Abstract

The study reported the development of FFBPNN architecture and its corresponding software to predict the rice
production data for three seasons in 31 districts of Tamilnadu. It was found that the training and testing data were
exactly matching with the predicted data. It was also found that the Absolute Relative Error (ARE) was found to be
zero at the 9th iteration itself. The FFBPNN system was improved by integrating it with the best fitting models using
the curve expert software. The improved FFBPNN with best fitting model was used to predict the area of rice and its
production. The predicted data was compared with the observed data. The paired t-test was conducted between the
observed and predicted data. It was found that there is 67% of fittings are showing insignificant difference between
the observed area of rice and predicted area of rice cultivation. Similar test was also conducted for the rice
production data; it was found that there is 73.3% of fittings showing insignificant difference between the observed
and predicted data.