What question did this study set out to answer?

This work evaluates various regression algorithms to improve airfare pricing predictions.

May 7, 2026

Flight Price Prediction Using Machine Learning and Deep Learning: A Comparative Study

Key Points

This work evaluates various regression algorithms to improve airfare pricing predictions.
Compared eleven regression algorithms from machine learning and deep learning.
Classical models included Linear Regression, Random Forest, and Gradient Boosting.
Evaluated models using metrics such as R², MAE, and RMSE.
TabTransformer and ExtraTreesRegressor achieved R² values above 0.99.
Machine learning models outperformed traditional approaches in airfare predictions.
Deep learning models demonstrated competitive performance using complex architectures.

Abstract

Airfare pricing is a highly dynamic and complex phenomenon influenced by numerous variables including departure time, number of stops, days to departure, flight class, and seasonal demand patterns. Accurate fare prediction offers practical value for cost-sensitive travelers and revenue-management optimization by airlines. This work presents a systematic comparative evaluation of eleven regression algorithms, spanning classical machine learning and contemporary deep learning approaches. Classical models include Linear Regression, Ridge, Lasso, Decision Tree, Random Forest, Extra Trees, Bagging, K-Nearest Neighbors, Gradient Boosting, and XGBoost. Five deep tabular architectures are benchmarked: MLP, DeepResNet1D, AttentionNet, WideAndDeep, and TabTransformer. Six CNN backbones (VGG11, VGG13, ResNet18, ResNet34, MobileNetV2, MobileNetV3) are also evaluated using synthetic 2-D image representations. All models are assessed across seven metrics: MAE, MSE, RMSE, R², Adjusted R², RMSLE, and MAPE. Results show that TabTransformer and ExtraTreesRegressor achieve R² exceeding 0.99.

Bookmark

Cite This Study

Dhanush et al. (Tue,) studied this question.

synapsesocial.com/papers/69fbefef164b5133a91a4033 https://doi.org/https://doi.org/10.64388/irev9i11-1717258

Bookmark