About

This project analyzes user interaction data from the Zerve platform to understand which behaviors and workflows predict successful long-term usage. The dataset contains more than 400,000 event records generated by over 5,000 users interacting with notebooks, AI agents, and workflow tools.

To study behavioral patterns, a user-event matrix was constructed representing more than 140 distinct event types. A success metric was defined by identifying the top 20% most active users based on overall interaction activity.

Using these features, a Random Forest machine learning model was trained to predict successful users from their behavioral patterns. The model achieved approximately 98% accuracy on the test dataset.

Feature importance analysis revealed that AI-assisted workflow actions such as agent tool calls, canvas summary requests, code execution events, and credit usage are the strongest predictors of successful platform usage.

These insights suggest that users who actively experiment with AI tools and automation workflows are more likely to become successful Zerve users. The findings can help guide product improvements that encourage experimentation and deeper engagement with AI-driven workflows.