Flight fare prediction using machine learning

Abstract

This paper deals with the forecast of Flight Price of domestic airlines.Revenue management relies heavily on forecasting. Air passengers (buyers) frequently search for the ideal time to buy tickets in order to save as much money as possible, whilst airlines (sellers) constantly strive to maximize their profits by adjusting different rates for the same service. For flying tickets, the airline uses dynamic pricing. Flight ticket costs fluctuate throughout the day, especially in the morning and evening. It also varies according to the holidays or festival season. The cost of a plane ticket is determined by a number of distinct factors. A lot of factors influence the cost of an airline ticket, including the location of source and destination, purchase time, number of stoppage, and so on. The sellers have all of the information they need (such as past sales, market demand, consumer profile, and behavior) to decide whether to raise or lower airfares at various times leading up to departure dates. Buyers, on the other hand, have limited access to information, which is insufficient to anticipate flight costs. It will offer the optimum time to buy the ticket based on parameters such as departure Date, Arrival Date, Source, Destination, Stoppage and Airline Name. To use Machine Learning (ML) models, features are retrieved from the gathered data. Then, using this data, we want to create a system that will assist consumers in deciding whether or not to purchase a ticket. Extracted features of a typical domestic flight of a year are taken as data and other conditions that may affect the flight is taken into consideration. The information is applied to machine learning models to predict flight ticket prices which uses the XGBoost algorithm that has given us 84.46% accuracy of prediction of the output price. We selected XGBoost as our chosen model after analyzing and visualizing 6 different Regressor models.