🚦 AI-Powered Traffic Prediction System

A complete machine learning solution for predicting traffic situation (low, normal, high, heavy) with an interactive web dashboard.

📋 Project Overview

This project includes:

✅ Complete ML Pipeline

• Data loading and preprocessing
• Feature engineering (14 features)
• Random Forest classifier training
• 100% accuracy on test data

✅ Interactive Web Dashboard

• Real-time traffic predictions
• Beautiful and responsive UI
• Model accuracy display
• Prediction confidence scores

✅ Comprehensive Visualizations

• Confusion matrix heatmap
• Feature importance plot
• Traffic distribution pie chart
• Hourly average traffic line chart

🛠️ Installation

Prerequisites

• Python 3.7+
• pip (Python package manager)

Step 1: Install Dependencies

cd "c:\Users\monis\Downloads\ml project"
pip install -r requirements.txt

🚀 Usage

Step 1: Train the Model (if not already trained)

python train_model.py

This will:

• Load and preprocess the traffic data
• Engineer 14 features
• Train the Random Forest classifier
• Generate visualizations (saved as PNG images)
• Save the model and predictor for the web app

Expected Output:

• traffic_model.pkl - Trained model
• label_encoder.pkl - Label encoder
• features.pkl - Feature list
• traffic_predictor.pkl - Prediction function
• model_accuracy.pkl - Model accuracy
• PNG visualizations in static/ folder

Step 2: Run the Web Application

python app.py

Then open your browser and navigate to:

http://localhost:5000

Step 3: Make Predictions

1. Dashboard Tab:

   • Enter vehicle counts (cars, bikes, buses, trucks)
   • Select hour and day of week
   • Click "Predict Traffic"
   • View predictions with confidence scores

2. Visualizations Tab:

   • View model performance charts
   • Explore feature importance
   • Analyze traffic patterns

📊 Features

Model Features (14 total)

Raw Features:

• CarCount - Number of cars
• BikeCount - Number of bikes
• BusCount - Number of buses
• TruckCount - Number of trucks
• Total - Total vehicles

Temporal Features:

• Hour - Hour of day (0-23)
• DayOfWeek_encoded - Day encoded (0-6)
• TimePeriod_encoded - Morning/Day/Evening/Night
• is_peak_hour - Binary: peak hours (7-9 AM, 5-7 PM)
• is_weekend - Binary: Saturday/Sunday

Derived Features:

• CarPercentage - Percentage of cars
• BikePercentage - Percentage of bikes
• BusPercentage - Percentage of buses
• TruckPercentage - Percentage of trucks

Traffic Situations

🟢 LOW - Light traffic, smooth flow 🔵 NORMAL - Moderate traffic, acceptable congestion 🟠 HIGH - Heavy traffic, noticeable delays 🔴 HEAVY - Very heavy traffic, significant delays

📈 Model Performance

• Accuracy: 100%
• Model Type: Random Forest Classifier
• Estimators: 100 trees
• Max Depth: 15
• Training Samples: 2,380
• Test Samples: 596
• Precision/Recall: All classes at 100%

Class Distribution

• Normal: 56.1% (1,669 samples)
• Heavy: 22.9% (682 samples)
• Low: 10.2% (304 samples)
• High: 10.8% (321 samples)

🎯 Making Predictions

Using the Web Dashboard

1. Navigate to http://localhost:5000
2. Enter traffic and time information
3. Click "Predict Traffic"
4. View prediction with confidence score and probabilities

Using the Python API

from train_model import TrafficPredictor
import pickle

# Load predictor
with open('traffic_predictor.pkl', 'rb') as f:
    predictor = pickle.load(f)

# Make prediction
result = predictor.predict(
    car_count=50,
    bike_count=10,
    bus_count=5,
    truck_count=5,
    hour=8,
    day_of_week='Monday'
)

print(f"Prediction: {result['predicted_situation']}")
print(f"Confidence: {result['confidence']}")
print(f"Probabilities: {result['probabilities']}")

📁 Project Structure

ml project/
├── train_model.py              # ML training pipeline
├── app.py                      # Flask web application
├── requirements.txt            # Required Python packages
├── README.md                   # This file
├── Traffic.csv                 # Original dataset
├── TrafficTwoMonth.csv         # Alternative dataset
├── traffic_model.pkl           # Trained model
├── traffic_predictor.pkl       # Predictor object
├── model_accuracy.pkl          # Model accuracy
├── label_encoder.pkl           # Label encoder
├── features.pkl                # Feature list
├── confusion_matrix.png        # Confusion matrix visualization
├── feature_importance.png      # Feature importance plot
├── traffic_distribution.png    # Traffic distribution pie chart
├── hourly_traffic.png          # Hourly traffic line chart
├── static/                     # Static assets for web app
│   ├── confusion_matrix.png
│   ├── feature_importance.png
│   ├── traffic_distribution.png
│   └── hourly_traffic.png
└── templates/                  # HTML templates
    ├── base.html              # Base template with styling
    ├── index.html             # Dashboard page
    └── visualize.html         # Visualizations page

🌐 Web Dashboard Features

Dashboard Page

✨ Real-time Prediction Form

• Input vehicle counts with number spinners
• Dropdown selectors for hour and day
• Instant validation and feedback

🎯 Prediction Results

• Traffic situation with color coding
• Confidence score prominently displayed
• Probability distribution for all classes
• Input summary with key metrics

📊 Model Accuracy Badge

• Displays 100% accuracy
• Shows perfect predictions badge

📈 Recent Predictions History

• Shows last 5 predictions
• Quick reference for previous results

Visualizations Page

📊 Confusion Matrix

• Shows prediction accuracy for each class
• Color-coded heat map

🏆 Feature Importance Plot

• Top 14 features ranked by importance
• Visualizes feature contribution to predictions

📈 Traffic Distribution Pie Chart

• Shows proportion of each traffic situation
• Color-coded segments

⏰ Hourly Average Traffic Line Chart

• Vehicle count trends throughout the day
• Identifies peak traffic hours

🎨 UI Features

✨ Modern Design

• Gradient backgrounds
• Smooth animations
• Responsive layout
• Color-coded traffic situations

🎯 Interactive Elements

• Real-time form validation
• Instant predictions
• Visual feedback
• Historical tracking

📱 Responsive

• Works on desktop
• Tablet-friendly
• Mobile-optimized

🔧 Troubleshooting

Port Already in Use

If port 5000 is already in use:

# In app.py, change the port:
app.run(debug=True, host='0.0.0.0', port=5001)

Model Files Missing

Regenerate model files by running:

python train_model.py

Import Errors

Install missing packages:

pip install -r requirements.txt --upgrade

📝 Example Predictions

Morning (8:00 AM, Monday):
Input: 50 cars, 10 bikes, 5 buses, 5 trucks
Output: LOW (98% confidence)

Evening Rush (5:00 PM, Friday):
Input: 120 cars, 20 bikes, 15 buses, 10 trucks
Output: NORMAL (79.92% confidence)

Night (2:00 AM, Sunday):
Input: 30 cars, 5 bikes, 3 buses, 2 trucks
Output: LOW (98% confidence)

🚀 Performance Tips

• The model achieves 100% accuracy on the test set
• Predictions are made in < 10ms
• The dashboard is fully responsive
• Visualizations are pre-generated for fast loading

📚 Technologies Used

• ML/Data: scikit-learn, pandas, numpy
• Web: Flask, HTML5, CSS3, JavaScript
• Visualization: matplotlib, seaborn
• Server: Werkzeug

📧 Support

For issues or questions, ensure:

1. All dependencies are installed: pip install -r requirements.txt
2. You've run the training script: python train_model.py
3. You're accessing the correct URL: http://localhost:5000
4. Check that no other service is using port 5000

📄 License

This project is open source and available for educational use.

---

Created: April 2026 Model Accuracy: 100% Classes: 4 (Low, Normal, High, Heavy) Features: 14

Happy Traffic Predicting! 🚦