The project demonstrates the use of the PACE workflow (Plan → Analyze → Construct → Execute) to design a project proposal and strategy for building a machine learning model that classifies TikTok content as either a claim or an opinion.

TikTok’s platform receives an overwhelming number of user-reported videos and comments that may contain claims. Reviewing and moderating this content manually creates significant bottlenecks. The goal of this project is to:

Build a predictive classification model to determine if a video contains a claim.

Help TikTok reduce its content moderation backlog.

Improve user safety, trust, and content quality through better prioritization.

This project represents the planning stage of a larger data science effort at TikTok to create scalable moderation tools and enhance platform safety.

Defined scope and deliverables of the classification project.

Identified project stakeholders (data science team, operations, finance, executives).

Outlined milestones for data preparation, modeling, and reporting.

Selected Python as the primary environment for model development.

Inspected TikTok’s dataset for missing or inconsistent values.

Performed exploratory data analysis (EDA) to uncover data trends.

Identified features most likely to contribute to classification accuracy.

Planned hypothesis tests to validate assumptions.

Designed and tested regression models for classification.

Built a machine learning pipeline for claim detection.

Checked assumptions of regression modeling to ensure accuracy.

Created visualizations to communicate findings to executives.

Validated and refined the final classification model.

Produced presentation materials for leadership.

Delivered a clear, actionable project proposal to TikTok executives.

PACE Strategy Executive Summaries.

Cleaned dataset and full EDA report.

Classification models and validation metrics.

Stakeholder-ready presentation materials and insights.