May 2022 – Jun 2022 · 1 mo · Kelowna, British Columbia, Canada · Remote

• Investigated the best Machine Learning Models for estimating agricultural yield based on growing patterns at the township level within the Canadian Prairie domain. Evaluated the effectiveness of several Machine Learning Algorithms in the prediction of canola crop yield and compared the outcomes with the observed values, using the Canadian Prairie domain as the research area. The project employs machine learning techniques and algorithms using the Python programming language and its numerous libraries and packages. Machine Learning algorithms such as Ridge Regression and LASSO are employed for regularization, while Principal Component Analysis (PCA) aids in dimensionality reduction. Deep learning algorithms like Feed Forward Neural Network and MLP Regressor are also utilized. Results indicate that higher accuracy and lower test Mean Squared Error (MSE) are achieved with a higher number of townships per eco-district, and the accuracy zone aligns with areas of high Canola density. Notably, machine learning methods outperform deep learning in modeling with a low number of records. This project seeks to provide our client, Agriculture and Agri-Food Canada, with an optimal model to enhance predictions of future crop yields, thereby aiding in climate change mitigation and agricultural planning.