COCONUT YIELD PREDICTION USING BACK PROPAGATION AND SPLIT ALGORITHM IN BAY ISLANDS

Abstract

The Andaman and Nicobar Islands comprise chain of more islands in addition to a number of islets and rock outcrops in the Bay of Bengal. Coconut is the one of the major crops in these islands. The main focus of the study is to investigate a neural network has been used to coconut yield prediction in Andaman and Nicobar Islands using with weather parameters. Data and information relating to coconut yield from CARI research farm has been collected for 40 years. Weather data such as average yearly rainfall; average mean temperature, Relative humidity, wind speed, evaporation and sunshine hours of relevant period 1980 to 2018 have been also obtained. In this study back-propagation and split algorithm has been used to get accurate yield prediction. The network was trained using 22 patterns each of 9 inputs. In this study split network configuration is used. It is well known, that the number of nodes in the hidden layer of the network should be such that the nodes represent the patterns. As the number of training patterns increased, the number of nodes in the hidden layer would also increase. As the number of nodes increased, the size of the network would increase. The developed ANN model used the split algorithm concept and the processing data will result in better forecasting.