DSC723 - Flight Delays (Hanif)

Phase 2 - Data Understanding

Phase 3 - Data Preparation

Phase 1 - Business Understanding

• Domain: US commercial aviation operations (2008 data).

• Business problem: Arrival delays cause high operating costs, passenger dissatisfaction and penalties. Airlines and airports need to predict at-risk flights early.

• Analytical question: Can we predict whether a flight will arrive significantly late (ArrDelay ≥ 15 min), and what are the main drivers of the delay?

• Target (label): a new column DelayClass = Late if ArrDelay ≥ 15, otherwise OnTime.

• Task type: Binary Classification.

• Success criteria: high F1-score & ROC AUC on the test set; actionable insight.

• Clearly state why DepDelay and the delay-cause columns are removed (to avoid data leakage) — this is the deeper understanding the rubric rewards.

Phase 4 - Modeling

Phase 5 - Evaluation

Decision Tree

Random Forest

Gradient Boosted Tree

Logistic Regression

Naive Bayes

Phase 6 - Visualisation

Check Imbalanced Data

Naive Bayes

Decision Tree

Random Forest

Gradient Boosted Tree

Logistic Regression