Improved Deep Learning Based Prediction of Crop Yield Using Bidirectional Long Short Term Memory

Abstract

Yield estimation for crops is an important problem to be addressed with respect to agricultural planning. Deep learning and machine learning provide feasible solutions for the above issue. In the current research perspective yield prediction is vastly addressed and more techniques evolve in the objective to improve accuracy. This research work tries to analyze the predictive accuracy of machine learning methods along with deep learning methods. The models are analyzed by metrics such as the Mean absolute percentage error (MAPE), root-mean-square error (RMSE), mean absolute error (MAE), and accuracy. The proposed Bidirectional LSTM deep learning technique exhibits high accuracy and low error rates in order to predict rice yield in INDIA The results show that Bi-LSTM have better accuracy when compared to other models, Bi-LSTM and XGB obtain the lowest RMSE errors, the lowest MAPE error percentage is for Bi-LSTM and the lowest MAE errors is also for Bi-LSTM. Since Bi-LSTM exhibits crop yield model with better accuracy and lowest errors it is suitable massive crop yield prediction in agricultural planning.