Deep Learning Model For Predicting Sorghum Yield: A Case Of Kisumu County

Abstract

Agriculture is said to be the backbone of Kenya’s economy contributing to over 20% of the country’s Gross Domestic Product (GDP). More than 40% of the country’s population are employed by the agricultural sector and an estimated 70% of the rural population rely on agriculture. Agricultural productivity is however dwindling owing to climate change related risks such as longer drought periods. In an effort to ensure sustainability and food security, different strategies are being implemented like climate smart agriculture which advocates for increased agricultural productivity through sustainability. Crop yield forecasting is one of the ways which can help provide useful information to policy makers and scientists to come up with sustainable agricultural strategies. It will also help farmers make informed farming decisions. Crop yield prediction is however a difficult task since many factors are considered when coming up with the ideal set of independent variables. Many studies have been conducted on predicting different crops yield using machine learning algorithms and different factors depending on the availability of data and the scope of the research. The main objective of this thesis is to come up with a deep learning model that predicts sorghum yield in Kisumu County. Deep learning is a preferred choice of machine learning algorithms because of its ability to have multiple hidden layers which increases the accuracy levels. The model will try an all-inclusive approach where all factors affecting sorghum yield production will be considered like environmental variables, agronomic, social and economic variables. Historical data obtained from the KALRO data portal will be used in this study. The Root Mean Squared Error (RMSE) and Mean Squared Error (MSE) will be used to evaluate the prediction performance of the model.