LSTM Traffic Flow Predictor

📋 Project Overview

Predictive Modelling service for the Athlone "Orange Loop" traffic optimization system.

Goal: Use Long Short-Term Memory (LSTM) neural networks to forecast vehicle flow 15 minutes into the future with Mean Absolute Error (MAE) < 10%.

Why LSTM?

• Learns temporal patterns (e.g., morning rush 8:15-9:00am)
• Handles non-linear traffic spikes (sudden congestion)
• Captures "memory" of traffic events across hours
• Superior to ARIMA for complex urban traffic

---

📊 Data Flow

SUMO Simulation
    ↓
edgeData.xml (vehicle counts, speed, occupancy per junction)
    ↓
Data Loader (extract hourly vehicle flow)
    ↓
Preprocessor (scale, normalize, create sliding windows)
    ↓
LSTM Model (trained on historical patterns)
    ↓
Forecast (predicted vehicle flow 15 min ahead)
    ↓
RL Inference Service (uses forecast for signal timing)
    ↓
Traffic Signal Control

---

🔑 Key Concepts

What is LSTM?

A type of neural network that:

• Remembers long-term patterns (e.g., "Mondays are always congested 8-9am with School / Work traffic")
• Forgets irrelevant old events (e.g., "An accident that happend 3 hours ago is now irrelevant")
• Learns non-linear relationships (e.g., "When School Reopens, rush hour is 8:15-9:00, but on holidays it's 9:30-10:30")

LSTM vs ARIMA

Aspect	ARIMA	LSTM
Pattern Type	Linear trends	Non-linear spikes
Memory	Limited (p,d,q params)	Long-term via gates
Sudden Changes	Poor	Good
Rush Hour Patterns	Struggles	Excellent
Data Amount Needed	Small	Large (1000+)