🚕 Manhattan Taxi Demand Prediction

A sophisticated forecasting system for predicting taxi demand across Manhattan using a hybrid approach combining machine learning and deep neural networks. The system uses a CNN-LSTM Encoder-Decoder model for Manhattan's top 5 locations and a Stacked Machine Learning model for remaining locations.

📑 Table of Contents

• Overview
• Features
• Data Sources
• Architecture
• Models
• Usage
• Results
• Technologies Used

🎯 Overview

This project develops a comprehensive taxi demand forecasting system for Manhattan, leveraging historical trip data from 2020-2022. The system combines advanced deep learning techniques with traditional machine learning approaches to provide accurate predictions for different areas of Manhattan.

⭐ Features

• 🔄 Real-time demand prediction visualization
• 🗺️ Interactive heatmap of Manhattan zones
• 📊 Hourly and daily demand pattern analysis
• ⛈️ Weather impact analysis
• 🎊 Holiday and weekend demand patterns
• 📍 Top locations analysis
• 📱 Detailed analytics dashboard

📚 Data Sources

• 🚖 Taxi trip records from NYC TLC (2020-2022)
• 🌤️ Weather data from Visual Crossing
• 🗽 NYC taxi zone information
• 🎉 US holiday data

🏗️ Architecture

The system uses a hybrid architecture:

• 🧠 CNN-LSTM Encoder-Decoder: For top 5 Manhattan locations
• 📚 Stacked Machine Learning Model: For remaining locations
• ☁️ AWS EMR Spark Cluster: For data processing
• 💻 Streamlit: For web application interface

🤖 Models

CNN-LSTM Encoder-Decoder

• 📈 Performance Metrics:
  • R-Square: 95.6%
  • Mean Square Error: 839.1
  • Mean Absolute Error: 19.56
  • Root Mean Square Error: 28.96

Stacked Model

• 📊 Performance Metrics:
  • R-Square: 96.72%
  • Mean Square Error: 125.51
  • Mean Absolute Error: 5.82
  • Root Mean Square Error: 11.2

🎮 Usage

• 📊 Dashboard View
  • View real-time demand predictions
  • Analyze demand patterns through interactive maps
  • Monitor key metrics and trends
• 📈 Detailed Analysis
  • Explore time series analysis
  • Compare demand across locations
  • Analyze weather impact

🎯 Results

• ✅ Successfully predicted taxi demand with over 96% accuracy
• 🔍 Identified key patterns in demand based on:
  • ⏰ Time of day
  • 🌦️ Weather conditions
  • 🎉 Special events
  • 📍 Location characteristics

🛠️ Technologies Used

• 🐍 Python
• 🧠 TensorFlow
• 📊 Scikit-learn
• 🌐 Streamlit
• ⚡ PySpark
• ☁️ Google Cloud Platform