✈️ Flight Delay Predictor – Data-Driven Analysis & Prediction

A real-world project using a 1.9M-row U.S. flight dataset to explore delay patterns, identify key causes, and build a predictive system using Random Forest — now deployed with an interactive Streamlit app.

📊 Objectives

• Analyze delays by hour, weekday, month, airline, airport, and distance
• Identify top delay causes for most delayed airlines
• Map airport and airline codes to readable names
• Build and deploy a Random Forest model to predict total delay

🔍 Key Insights

• Flights after 5 PM tend to have higher delays due to cascading schedule shifts.

• JetBlue Airways had the highest average delays, mostly from late-arriving aircraft.

• Peak delays in December and July (likely holiday-related); lowest in September–October.

• Regional airports like CMX, ACY, and PLN showed unexpectedly high average delays.

• Final Random Forest model achieved:

  • ✅ R² Score: 0.969
  • ✅ RMSE: ~17.7 minutes

🤖 Live App (Streamlit)

Predict delay in real time based on flight input:

🟢 Try the app

📷 Screenshot


---

🛠️ Tools & Technologies

• Python: pandas, seaborn, scikit-learn, matplotlib
• ML Model: Random Forest Regressor
• Web App: Streamlit
• Persistence: joblib
• Notebook: Jupyter for full analysis & modeling

---

🧪 Files in This Repo

File	Description
Flight_Delay_Analysis.ipynb	Full EDA, modeling, evaluation
app.py	Streamlit app source
flight_delay_rf_model.pkl	Trained model (joblib)
model_features.pkl	Feature list for safe prediction
README.md	You’re reading it
streamlit app screenshot.png	Streamlit UI preview (optional)

---

🔗 Dataset

• Source: U.S. Department of Transportation
• 📦 DelayedFlights.csv

---

📌 Future Work

• Improve model with additional features (e.g., weather, holiday flag)
• Deploy on cloud (Streamlit Cloud / AWS)
• Add airline-specific forecasting modules
• Option to explain delay causes based on input (e.g., Late Aircraft vs Weather)

---

🧠 Author

Kulsoom Banu Shaik GitHub • LinkedIn