Wheat yield prediction: Artificial neural network based approach

Wheat yield prediction modeling is an important area of study because of its potential contribution to food security since it may be perceived to be a good indicator for global food availability. Many studies have been conducted in order to determine the best models for wheat yield prediction using various types of data which are available; these models include CERES-Wheat model, SIRIUS model and AFRCWHEAT2 model. In this study, our wheat yield prediction model is designed using a Multi-Layer Perceptron (MLP) backpropagation-based- feed forward artificial neural network (ANN). The data used was weather data including: sun, frost, rain and temperature as the input parameters from year 1997-2007. The output parameter of the model is using the wheat yield data for the years 1997-2007. The data is divided into three separate sets; - for training, validation and testing. Our MLP was able to predict, wheat yield with an accuracy of 98 %. Hence our MLP based wheat yield prediction model shows great promise as a tool which will be able to provide relatively accurate wheat yield prediction and may be applied to other crops.