Comparison and Optimization Strategies of Airbnb Rental Prediction Models: An Empirical Study Based on Linear Regression, XGBoost and Random Forest

As a leading short-term rental platform in the sharing economy, Airbnb employs dynamic pricing based on multiple factors, posing challenges for accurate rental price prediction. This study aims to improve rental price prediction by comparing and optimizing three machine learning modelslinear regression, random forest, and XGBoostusing Airbnb listing data. We develop a multi-model prediction framework that exploits each models strengths: linear regression provides interpretability of key features, random forest captures nonlinear interactions, and XGBoost applies gradient-boosting techniques to minimize error. Using standard regression metrics (e.g., RMSE, MAPE) for evaluation, we find that XGBoost delivers the highest predictive accuracy (8% error), outperforming the random forest and linear regression models. The results indicate that XGBoosts enhanced ability to model complex market dynamics yields more reliable price estimates, whereas the baseline linear model and random forest show higher errors and signs of overfitting. In conclusion, our comparative analysis offers practical insights for stakeholders: hosts can leverage the improved model for optimal pricing strategies to maximize revenue and occupancy rates, the Airbnb platform can refine its smart pricing tool for greater market efficiency, and overall data-driven pricing strategies benefit the sharing economy by aligning stakeholder interests.