Spice Yield Prediction for Sustainable Food Production Using Neural Networks

Abstract

AbstractThe world population is increasing rapidly, and the consumption pattern of mankind has made a drastic drift over the recent years. Sustainable food production is important for our existence. The main focus of the study is to build a model that can predict the crop yield for spices such as black pepper, dry ginger, and turmeric based on given factors such as the district of cultivation, year of cultivation, area of production, production per year, temperature, and rainfall. The dataset was obtained from the Spice Board of India and Meteorological Database of India. The region primarily focused on is the districts of Kerala. Neural networks were used for the prediction, and a comparative study was done on different models such as deep neural network (DNN), recurrent neural network (RNN), gradient recurrent unit (GRU), long short-term memory (LSTM), bi directional long short-term memory (BiLSTM), backpropagation neural network (BPNN). The validation techniques taken into consideration include normalized mean absolute error (MAE), normalized root mean square error (RMSE), and mean absolute percentage error (MAPE). For dry ginger, GRU performed better compared to other algorithms followed by SRN. For black pepper, DNN performed better compared to other algorithms followed by simple recurrent network (SRN). For turmeric, GRU performed better compared to other algorithms followed by BPNN.KeywordsSpice yieldNeural networksFood productionComparative studyDeep learning