Recurrent Neural Network Based Forecasting of Crop Production in Nepal
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Pulchowk Campus
Abstract
In this study, Gated Recurrent Unit (GRU) model was used for predicting Rice crop
production in Nepal using climatic and fertilizer variables. The climatic variables used
were Maximum Temperature, Minimum Temperature, Morning Humidity, Evening
Humidity and Rainfall and Ecological Regions and fertilizer variables were Nitrogen,
Phosphorous, Potassium and Compost. When the model was trained on 70% of data
and tested on 30% of the data, the accuracy of the model was 81% for predicting the
production. When tested on year 2016, accuracy of the model was 81.33% and for
year 2017, the accuracy of the model was 73.33%. While GRU was compared with
baseline Artificial Neural Network (ANN) with same architecture for Siraha district, it
performed better than baseline ANN. But when input variables were increased, it
performed even better. This proved that GRU can be used for optimal prediction of
Rice crop in Nepal.
Description
In this study, Gated Recurrent Unit (GRU) model was used for predicting Rice crop
production in Nepal using climatic and fertilizer variables.