About

This canvas is a comprehensive user success prediction and engagement platform that trains machine learning models on 5,410 users' behavioral data to identify at-risk users and design re-engagement strategies. Starting with event-level feature engineering (hackathon_submission_overview), it builds a Gradient Boosting classifier achieving 97.3% ROC-AUC to score users by success risk, then segments the population into behavioral cohorts and performs temporal validation to prove the model predicts future behavior. Downstream, it executes a complete A/B test framework with power analysis and outcome simulation to evaluate targeted re-engagement campaigns on high/medium-risk users, alongside detailed segmentation analytics with visualizations that expose key behavioral patterns and feature adoption gaps between successful and at-risk users.