gkrowan/Rwanda-landslides

Data Science Clinic Winter 2025 # 2025-winter-rwanda-landslides ## Project Background ### Partner Organization: **University of Rwanda** ### Mentors: - **Tim Hannifan**, Assistant Clinic Director, External Mentor - **Ganghua Wang**, Post-Doctoral Scholar at UChicago, Internal Mentor - **Gayathri Jayaraman**, MSCAPP ‘25, TA ### Team members: - **Rachelle Cho**, UChicago Undergraduate, Class of ‘25, rcho@uchicago.edu - **Jenny Li**, UChicago Undergraduate, Class of ‘25, jennyyueli@uchicago.edu - **Grace Rowan**, UChicago Undergraduate, Class of ‘25, gkrowan@uchicago.edu - **Sana Fessuh**, UChicago Undergraduate, Class of ‘25, shfessuh@uchicago.edu ### Project Background: Landslides pose serious risks in Rwanda, particularly in the face of changing climactic conditions. However, research regarding predictive efforts to mitigate landslides remains limited. This project focuses on the Gitwe-Kadhua Corridor, an area identified as high risk. Through the creation of various predictive models, we analyzed key environmental and topographic factors that may be contributing to landslide susceptibility in the region. We hope that our findings will contribute to enhancing early warning systems and inform mitigation strategies for landslides in Rwanda. ## Project Goals Our goal is to explore geotopical factors impacting landslide risk in Rwanda's Gitwe-Kaduha Corridor and assess risk through various predictive frameworks that encorporate the relevant features identified in our data. ### Preparation & Visualization of Current Data We explore and visualize the raw data provided by the University of Rwanda. We devise a workflow for merging files to create a complete dataset of various environmental factors across the Gitwe-Kaduha Corridor. ### Data Anlysis and Modeling We used 4 different types of classification frameworks to predict landslide risk from the available features: an Ordered Linear Model, a Random Forest Model, a Neural Network, and a Large Language Model. ### White Paper In a white paper, we …