DS785-Capstone

Repository containing code for a Masters Thesis analyzing causes of airline delays

Primary Scripts: Ran in this Order

1. Data pre-processing:

• Joins and cleans airline and weather datasets
• Drops unnescessary and redundant variables
• Outputs combined CSV file for use in Time Series Forecast

2. time_series_forecast:

• Uses LSTM model to forecast % of flights delayed from each airplane 24 hours in advance, adds this forecast to the dataset
• Calculates airport level capacity metrics and adds these to dataset
• Outputs CSV file for use in Airline Modeling Script

3. Airline_modeling:

• Performs encoding on categorical variables
• Trains models in 2-hour and 24-hour prediction scenarios
• Assesses feature importance of each model in each scenario
• Visualizes model training and feature importance for use in presentation

Secondary Scripts: For generating some of the visualizations in the project

1. Time Series EDA: Produces time series visuals showing on time flight performance over a number of time measures(week, month, etc.)