Forecasting of the rain-fed maize yield in Tanzania using machine learning
Loading...
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Sokoine University of Agriculture
Abstract
Yield monitoring is vital for food security in the country. The uncertainty that may
influence fluctuations in yield is impeccable to the nation's food security strategy.
Due to climate change around the world, yield fluctuation has dramatically been
affected and led to food supply shortage in most developing countries, including
Tanzania, since most of the staple food depends on rainfed agriculture, which is hit
by high temperatures and variations in rainfall. Bad weather also contributes to
diseases, pests and weeds which greatly challenge the growth of crops, particularly
the top grain and maize. Most countries have adopted several yield prediction
methods to mitigate the effects to resolve the situation. Most countries have shown
that emerging techniques such as machine learning provide good predictions to help
countries mitigate the problem and secure food security.
Machine learning regression models (linear, AdaBoost, gradient boosting, k-
Neighbour, random forest and stacking ensemble) are trained and evaluated using
dataset obtained from the online database. Normalized Difference Vegetation Index
(NDV1) is used to predict vegetation that are later assumed to locate probability of
maize fields. The climate data from those areas is then subjected to training on
forecasting maize yield.
Therefore, the results have shown that Machine learning methods like a stacking
ensemble, which combine several other models and use Random Forest as the final
model achieved low Mean Absolute Error (MAE), Mean Squared Error (MSE) and
Root Mean Squared Error (RMSE) of 0.27, 0.12 and 0.34 Ton per Ha for each
district respectively.
The results suggest machine learning model like stacking ensemble can be used by
policy makers and farmers to mitigate the effects of climate change in yields for a
particular season.
Description
Thesis
Keywords
Machine learning, Maize yield, Rain-fed