An Optimal Bidirectional Gated Recurrent Neural Network Model for Crop Yield Prediction

Abstract

Crop yield forecasting acts as a vital part of farm management, domestic food supply, global food trade, ecological sustainability, etc. Several features which affect the crop yield are plantation region, competence of irrigation system, variation in rainfall and temperature, soil quality and fertilization, and diseases. An automated crop yield prediction model using machine learning and deep learning approaches is necessary for effective decision making in agriculture. In this view, this study designs an optimal bidirectional gated recurrent neural network (OBGRNN) based crop yield prediction model. The goal of the OBGRNN technique is to forecast the crop yield using histoiical farming data. The OBGRNN technique performs two processes namely prediction and parameter optimization. At the initial stage, the BGRNN technique is applied for the actual prediction of the crop yield. Besides, in the next stage, the Differential Evolution (DE) is applied for the effectual tuning of the parameters involved in the BGRNN technique. The OBGRNN technique involves two processes namely prediction and parameter optimization and helps to improve the performance. A series of simulations were performed on the benchmark test dataset and the experimental values reported the better performance over the other techniques.