Inspiration
We wanted to investigate the lowest airline prices in order to determine the best predictors and how consumers can use these factors to understand trends with purchasing airline tickets.
What it does
Our model predicts airline price for the years recorded to have the lowest airline prices (2002, 2004, 2005, 2009, and 2020) for the 3 US airports that have the largest number of outgoing passengers: Boston Logan International Airport(BOS), Los Angeles International Airport (LAX), and Chicago O'Hare International Airport (ORD).
How we built it
We used tableau to create inital visualization for our exploratory data analysis. We used python code to build a random forest regression model and a gradient boosting regression model. We utilized these models to determine the top features used in providing the best prediction of airline fares in the years that were recorded to have the lowest prices.
Challenges we ran into
We had to experiment with using different regression models including lasso, random forest regression, and gradient booster regression models. We then calculated the error for each and finally chose the gradient booster regression model as the model that best fit the complexity of the dataset.
Accomplishments that we're proud of
We were able to apply the models by ranking the factors used in predicting fares by their coefficients and determining the top factors that provided the best prediction.
What we learned
We learned about different models that can be used for predictive analysis and how we can determine the best model to fit a dataset. We are also able to understand the methodology in determining the best predictors given a dataset with many independent variables for a dependent variable.
What's next for Comparing best predictors of Lowest Airfare Price
We used mean airline prices in this analysis but we can also use median in a future analysis to see if there is any difference. We can also use the model to predict airline fares for years with the lowest fares ranked below the top 5 to see how well the model performs.
Log in or sign up for Devpost to join the conversation.