✈️ Flight Delay Prediction using MLP

📌 Overview

This project predicts whether a flight will be delayed based on various factors such as weather conditions, holidays, and weekends.
The model is built using a Multi-Layer Perceptron (MLP) neural network with Softmax activation for binary classification.

📊 Dataset

This dataset contains real flight data from Air Canada flights operating:

• YYT (St. John's) → YYZ (Toronto Pearson)
• YYT (St. John's) → YHZ (Halifax Stanfield)

Features:

• Flight conditions
  • Weather at departure & arrival airports
• Temporal features
  • Whether the flight is on a weekend
  • Whether the flight is on a holiday
• Delay information
  • Flight delay in minutes (Delay (min))
• Target Variable:
  • Is Delayed → Binary classification (0: Not Delayed, 1: Delayed)

🏗 Model Architecture

The model consists of:

• Input Layer: Takes the flight features (weather, weekend, holiday, delay)
• Hidden Layers:
  • Dense (32 neurons, ReLU)
  • Dense (16 neurons, ReLU)
• Output Layer:
  • Dense (2 neurons, Softmax activation)
  • Predicts the probability of delay vs. no delay

📈 Model Performance

Metric	Value
Accuracy	0.7978
Precision	0.7250
Recall	0.5370
F1 Score	0.6170
AUC Score	0.7310

📝 Interpretation:

• The model achieves ~80% accuracy, meaning it correctly predicts delays in most cases.
• Precision (0.7250) suggests that 72.5% of predicted delays are correct.
• Recall (0.5370) indicates that only 53.7% of actual delays were identified, meaning there are still some missed delays.
• AUC Score (0.7310) shows that the model has a good ability to distinguish between delayed and non-delayed flights.

🚀 Installation & Usage

1️⃣ Install Dependencies

Ensure you have Python 3.8+, then install the required packages:

pip install -r requirements.txt

conda activate tf-env

conda deactivate