About

This canvas implements an end-to-end machine learning pipeline that predicts early-stage Power Users on the Zerve platform by analyzing their first 30 days of activity, combining event labeling, feature engineering, imbalanced classification, behavioral clustering, and inference demonstration. The workflow chains together exploratory data analysis, feature extraction across 7 and 14-day windows, dual model training (Logistic Regression and Gradient Boosting), user segmentation, and a live inference demo—all executed sequentially to identify and understand high-value users from raw event logs.