Strava Kudos Predictor
About this project
Created a tool that predicts kudos, a proxy for user interaction, on Strava activities (RMSE: 9.41) to see if it was random or if different attributes impact kudos in different ways. Attained over 4000 Strava activities using the Strava API and python. Engineered new features using domain knowledge. For example, features encapsulating different run types and times of day were added. Performed feature selection using a combination of SHAP values and feature importance. Optimized Linear (Lasso), Random Forest, and XGBoost Regressors using Optuna to reach the best model. Built an interactive API.
More projects by Jack Leitch
Recipe Recommendation System
Created a tool that recommends recipes based on ingredients inputted to help students eat better...
Customer Churn Prediction
Created an interpretable model that predicts churn on telecom data found on Kaggle (F1: 0.796)...