Airline Workflow

<p><strong>Data Quality Check</strong></p><ul><li><p>Imported airline stock data (AAL, DAL, LUV, UAL), jet fuel prices, and NYSE ARCA Index</p></li><li><p>Conducted data quality check using&nbsp;<strong>Statistics node</strong></p></li><li><p>Focused on identifying&nbsp;<strong>missing values and unique values</strong></p></li><li><p>Evaluated data completeness and consistency before&nbsp;<strong>preprocessing</strong></p></li></ul><p></p><p><strong>Data Preprocessing Overview</strong></p><ul><li><p><em>Stock and ARCA Prices</em></p></li></ul><ol><li><p>Cleaned dataset by removing duplicates and converting variables into numeric format</p></li><li><p>Standardized and cleaned datasets across&nbsp;<strong>consistent weekly base intervals</strong></p></li><li><p>Transformed price data into<strong>&nbsp;return-based</strong>&nbsp;metrics</p></li><li><p>Constructed<strong>&nbsp;volume-weighted average&nbsp;</strong>price and return relative to the market benchmark</p></li><li><p>Generated&nbsp;<strong>lagged variables&nbsp;</strong>to capture delayed market effects</p></li></ol><ul><li><p><em>Fuel Price</em></p></li></ul><ol><li><p>Cleaned dataset by removing duplicates and converting variables into numeric format</p></li><li><p>Transforming raw fuel price data into structured, lagged return-based features for time-series regression.</p></li><li><p>Then, do the -1,-2,-3,-4 lag accordingly ⟶&nbsp;<strong>capture delayed market reactions</strong></p><p></p></li></ol><p><strong>Regression Overview</strong></p><ul><li><p><em>Steps</em></p></li></ul><ol><li><p>Separated models into&nbsp;<strong>Linear</strong>&nbsp;and&nbsp;<strong>Exponential regression frameworks</strong></p></li><li><p>Linear model captures&nbsp;<strong>direct (additive) relationships</strong></p></li><li><p>Exponential model captures&nbsp;<strong>non-linear (multiplicative) effects</strong></p></li><li><p>Applied&nbsp;<strong>multiple lag structures (Lag 1–4)&nbsp;</strong>to test delayed impact</p></li><li><p>Compared model performance to identify the best predictive specification</p></li></ol><ul><li><p><em>Components</em></p></li></ul><ol><li><p>Each regression model is applied across three time periods</p></li><li><p>Full sample: 2019–2025</p></li><li><p>Sub-period&nbsp;<strong>Chow test analysis</strong>: Pre-war and Post-war (2022-8 break)</p></li><li><p>Allows comparison of<strong>&nbsp;model behavior</strong>&nbsp;under different market conditions</p><p></p></li></ol><p></p><p></p>