🚗 Uber Driver Activation Prediction

A machine learning project that predicts whether a new Uber driver will actually start driving — built for Soar into Data Datathon.

---

🚀 What We Built

We analyzed Uber’s driver signup data to identify key patterns and predict which users will become active drivers.

Our model helps Uber:

• Understand early drop-off points
• Improve driver onboarding
• Allocate support resources more effectively

---

📊 Key Results

• Models Used: Logistic Regression, Random Forest, XGBoost
• Best Accuracy: ~93%
• Best Balance of Precision/Recall: XGBoost (92% precision, 94% recall)
• Confusion Matrix Analysis: Highlighted importance of data balancing

---

🧹 Data Preparation

• Removed irrelevant columns (e.g., IDs, unused text)
• Handled missing values (e.g., 'unknown', -1)
• Created binary target: started_driving
• Engineered features like onboarding_delay and vehicle_age
• Balanced dataset: 6,000 drivers and 6,000 non-drivers

---

📁 Files

• Driven by Data.ipynb — full code and model development
• new_copy_combined.csv — cleaned dataset used for training
• README.md — this file

---

💡 What We Learned

• Accuracy isn’t enough — precision and recall revealed model weaknesses
• Balancing the data was critical to model fairness
• XGBoost provided the best trade-off between flexibility and performance

---

👥 Team

• Justin Le– ML / Data Cleaning
• Johnson Nguyen – ML / Data Cleaning
• Joe Cheung – Visualization / Presentation

---

📣 Built For

🛠️ Soar into Data Datathon | 📅 2025