Math Colloquium: Yew-Meng Koh, Mathematics Department

A study on farming practices and farm characteristics was carried out in three sub-counties of Kibaale District in Uganda. Using techniques from Statistical Learning, we identify salient features of these farms which are most useful for predicting farm incomes. We discuss the various models used, methods for comparing their prediction accuracy and interpretability, and conclude with a list of levels of features which lead to the highest predicted likelihood of high farm incomes.