Forecasting Wheat Yield Using Long Short- Term Memory Considering Soil and Metrological Parameters

Abstract

Early-season Crop Yield Prediction can assist farmers in India's leading economic sector of agriculture by assisting them in formulating their decision-making strategies. Deep Learning approaches have surpassed conventional statistical methods for yield prediction and crop forecasting as the artificial intelligence field has grown. The goal of the current work is to employ a LSTM model to estimate wheat crop yields in India. The dataset in this paper consist of soil and the metrological parameters. On the basis of consideration of individual factor one at a time, soil parameters such as temperature, humidity, moisture, soil type, crop, nitrogen, potassium, phosphorous in addition to nourishment used with consideration of metrological data, it contains minimum and maximum temperature as well as rainfall. At the end, we are able to get the accuracy and mean absolute error with R2 value for both the parameters. Later, we can merge these two parameters and get more efficient results for accurate prediction.