Rice Yield Estimation Using Deep Learning

Abstract

AbstractGlobal and regional food security heavily relies on effective yield estimation results. Thus precise and on-time rice yield estimate or prediction is a pivotal factor not only to ensure food security but also for sustainable development of agricultural resources. Machine learning and deep learning are proving to be exemplary support tools for decision making for rice yield estimation or prediction, such as selection of the rice varieties that need to be grown and also decisions involving the management of crops during growing season. Several researchers have put forth a variety of deep learning as well as machine learning algorithms that have helped estimate rice yield time and again. This paper proposes a LSTM based model to predict the Rice yield of the data collected for all 314 blocks of Odisha by ICAR - National Rice Research Institute (NRRI), Odisha. In this study, we get 0.07 RMSE score for training data and 0.21 RMSE score for test data. The model is also evaluated based on the various performance metrics for three rice datasets. The overall performance for the rice datasets is evaluated to be 0.989 recall, 0.979 precision, 0.989 accuracy and 0.984 F1 score.KeywordsMachine learningDeep learningRice yield estimationFood securitySustainable development