Predicting the Area and Production of Sugarcane in Tamil Nadu, India Using Neural Networks

Sugarcane is a major cash crop in India, grown in almost 5 million hectares with a production of 339 million tonnes. Tamil Nadu contributes significantly to the production of sugarcane. Data from the past year show a huge fluctuation in the area and production of sugarcane in the state. Predicting the area and production employing traditional modelling techniques fails because the assumptions are never attained in the field. To overcome this, soft computing techniques like artificial neural networks (ANNs) are used. In this study, a multilayer perceptron neural network (MLP-NN) with back-propagation was used to predict the area and production of sugarcane in Tamil Nadu. The MLP-NN (2,2) model predicts the area with minimum mean absolute error (MAE; 18.139) and root mean squared error (RMSE; 23.058) values and with high accuracy (99%). For production, the MLP-NN (2,1) model estimates minimum MAE (24.875) and RMSE (31.199) values with high accuracy (99%). So, MLP-NN (2,2) and MLP-NN (2,1) are the best ANN models to predict the area and production of sugarcane in Tamil Nadu respectively. Additionally, ANN models perform better in predicting nonlinear stochastic data.