🗽 NYC Taxi Ride Duration Prediction

This project is focused on predicting the duration of taxi rides in New York City using machine learning. It was developed as part of a data science learning track, using real-world data from the Kaggle NYC Taxi Trip Duration dataset.

📌 Objective

Build a regression model to predict the ride duration (in seconds) based on pickup and dropoff locations, times, and other engineered features.

---

📂 Project Structure

• Project-5._NY_taxi_ride_duration.ipynb
• .getignore
• README.md
• submission_gb.csv

The entire analysis and modeling process is contained in the notebook, which is divided into key sections:

• Data Loading and Exploration
• Feature Engineering
• Data Cleaning and Preprocessing
• Model Training and Evaluation
• Feature Importance
• Final Results and Submission Preparation

---

📊 Dataset Overview

The dataset includes over 1 million taxi trips with the following key features:

• pickup_datetime, dropoff_datetime
• pickup_longitude, pickup_latitude
• dropoff_longitude, dropoff_latitude
• passenger_count
• store_and_fwd_flag
• trip_duration (target)

---

🧪 Methodology

🔧 Feature Engineering

• Distance calculation using the haversine formula
• Datetime features (hour, weekday, month, etc.)
• Direction and speed estimates

🚀 Modeling

• Baseline: Linear Regression
• Advanced models:
  • Decision Tree
  • Random Forest Regressor
  • Polynomial Regression
  • Gradient Boosting (XGBoost)

🏆 Evaluation

• Metric: Root Mean Squared Log Error (RMSLE)
• Cross-validation used to avoid overfitting

---

🧠 Key Insights

• Datetime and geospatial features strongly influence ride duration.
• XGBoost outperformed other models with optimized hyperparameters.
• Feature importance analysis revealed trip distance and pickup hour as critical predictors.

---

✅ Final Model Performance

Model	RMSLE (CV)
Linear Regression	~0.59
Random Forest	~0.44
XGBoost (tuned)	~0.39

---

📎 Requirements

• Python 3.8+
• Jupyter Notebook
• pandas, numpy, matplotlib, seaborn
• scikit-learn
• xgboost