Climate change impacts on agricultural production: a machine learning approach for Greece

In the present thesis, we investigate the impacts of climate change on agricultural production. The first part is to acknowledge the consequences of climate change on environment and the problems that harm the agriculture sector. The goal is to understand the relationships between climate variables and agricultural Total Factor Productivity (TFP), which can be further used in predictive analysis to avoid possible disaster. Our approach focuses on linear models and machine learning techniques. We deploy linear regression, fixed effects, breakpoints regression, time series models, regression trees and boosted regression trees. In this way, we try to explain the variability of agricultural TFP, provide breakpoints analysis and determine threshold values of important variables. The applied models are checked for explainability, interpretability and predictive power, in order to evaluate which one is the most robust and effective, using well defined metrics. At last, we conclude to policy advices and technical suggestions for an applied use of our findings, along with possible future improvements.