Uber Trip Data Analysis & Classification

What needed to be solved

Turn an imperfect trip dataset into a defensible classification workflow with consistent evaluation across several models.

How the project addressed it

Prepared 1,156 records, explored trip patterns, engineered features, and evaluated four machine-learning models using accuracy and ROC-AUC.

My contribution

Built the workflow from cleaning and feature engineering through visualization, model comparison, and interpretation.

• Completed cleaning, exploratory analysis, feature engineering, and classification.
• Compared models using accuracy and ROC-AUC instead of a single metric.
• Presented the results with visualizations and a clear model-selection rationale.

What the work demonstrated

The strongest model achieved 95.2% accuracy and a 0.97 AUC score.