ML-Ensemble Major-Delay-Cause-Classifier

This repository presents a machine learning approach to classify the primary cause of flight delays at U.S. airports. The methodology emphasizes ensemble learning, extensive feature engineering, and class imbalance handling.

---

📄 Report

👉 [Click here to view the full PDF report](./Major_Delay_Cause for an airport.pdf)

📒 Notebook

👉 [Major_Delay_Cause.ipynb](./Major_Delay_Cause for an airport.ipynb)

---

Overview

• Task: Multi-class classification of delay causes: carrier, weather, NAS, late aircraft
• Dataset: 13,000+ aggregated monthly airport-carrier delay records
• Target Variable: Major_Delay_Cause (derived from delay components)

Methodology

• Preprocessing:
  • Missing value imputation (median)
  • Standardization of numerical features
  • One-hot encoding of categorical variables
• Feature Engineering:
  • Delay rates, risk scores, historical performance, seasonality indicators
  • Z-score normalization for airport/carrier
• Modeling:
  • Baselines: Logistic Regression, Decision Tree, KNN
  • Ensembles: Random Forest, Gradient Boosting, Bagging, AdaBoost
  • Final ensemble via Soft Voting on tuned Gradient Boosting and Random Forest

Performance

• Best F1 (Macro): 0.827 (Gradient Boosting with SMOTE)
• Voting Ensemble Accuracy: 88%
• Evaluation Metrics: Accuracy, Precision, Recall, F1-score, Confusion Matrix

Files

• Major_Delay_Cause.ipynb: Full pipeline including EDA, modeling, and evaluation
• exported_output.pdf: Complete project report with visualizations and analysis