🚕 Cab Fare Prediction: Machine Learning Approach

A complete end-to-end machine learning project that predicts cab fares based on ride features such as distance, time of day, weather conditions, and surge pricing. Built using real-world Uber and Lyft ride data from Boston, MA.

Author: Roshni Dodhi Date: October 2025 Project Type: Supervised Learning — Regression

---

📌 Problem Statement

Ride-hailing fares fluctuate based on a wide range of factors — distance, demand, time, weather, and service tier. This project builds and compares regression models to accurately predict the fare (price) for a given cab ride, helping both riders and companies better understand pricing dynamics.

---

📂 Dataset

Detail	Value
Records	693,071 rides
Features	57 columns
Target Variable	price (continuous)
Cab Services	Uber & Lyft
Location	Boston, MA (Nov–Dec 2018)

Key feature categories include ride metadata (source, destination, distance, cab type, service name), temporal features (hour, day, month), surge pricing, and a rich set of weather variables (temperature, humidity, wind, precipitation, cloud cover, visibility, etc.).

---

🛠️ Project Workflow

1. Data Loading & Initial Exploration — Shape inspection, data types, summary statistics, and missing value assessment.
2. Data Cleaning & Preprocessing — Handling ~55K null prices, type conversions, and preparing features for modeling.
3. Exploratory Data Analysis (EDA) — Visualizing price distributions, correlations, temporal patterns, and Uber vs. Lyft comparisons using Matplotlib, Seaborn, and Plotly.
4. Feature Engineering — Encoding categorical variables, scaling numerical features, and selecting the most predictive columns.
5. Model Building — Training three regression models: Linear Regression, Decision Tree, and Random Forest.
6. Model Evaluation — Comparing models using R², MAE, and RMSE on both train and test sets.
7. Conclusions & Business Recommendations

---

📊 Model Results

Model	R² Score	MAE	RMSE
Linear Regression	0.9347	$1.67	$2.24
Decision Tree	0.9960	$0.30	$0.55
Random Forest	0.9594	$1.28	$1.76

Best Model: Decision Tree — explains 99.6% of fare variance with an average prediction error of just $0.30.

---

💡 Key Insights

• Distance and surge multiplier are the strongest predictors of fare price.
• Peak demand occurs during morning (7–9 AM) and evening (5–7 PM) rush hours; late-night rides also carry higher fares.
• Weather impacts pricing — rain and poor conditions correlate with increased demand and surge pricing.
• Uber and Lyft show different pricing strategies, with premium tiers (Lux, Black) significantly more expensive than shared rides.

---

🔧 Tech Stack

• Language: Python 3.13
• Data: Pandas, NumPy
• Visualization: Matplotlib, Seaborn, Plotly
• Machine Learning: Scikit-learn (LinearRegression, DecisionTreeRegressor, RandomForestRegressor)
• Environment: Jupyter Notebook

---

🚀 Getting Started

# Clone the repo
git clone https://github.com/your-username/cab-fare-prediction.git
cd cab-fare-prediction

# Install dependencies
pip install pandas numpy matplotlib seaborn plotly scikit-learn

# Run the notebook
jupyter notebook Cab_Analysis_Final.ipynb

---

📁 Project Structure

cab-fare-prediction/
├── Cab_Analysis_Final.ipynb   # Main analysis notebook
├── README.md                  # Project documentation
└── data/                      # Dataset directory (add your CSV here)

---

🔮 Future Improvements

• Incorporate real-time traffic data
• Add event calendar features (sports games, concerts)
• Include driver availability and user ratings
• Experiment with deep learning and gradient boosting models (XGBoost, LightGBM)

---

📝 License

This project is for educational and portfolio purposes.