基于机器学习的机票价格预测研究

Machine Learning Based Flight Price Prediction

价格预测是经典的机器学习问题,过去已经有了不少相关研究。本研究与过往研究的不同在于,它们只在个别航线、数个月跨度的数据上做测试,不能覆盖机票业务的复杂性和周期性,而我们作为从业者,坐拥大范围长周期的数据,在数据上有巨大的优势。而且,我们长期工作在机票销售的一线,对机票价格的形成机制与变化趋势也有深刻的理解,这有助于我们更好地利用数据和构建模型。我们分别使用随机森林和XGBoost算法建立预测模型,均取得不错的效果。最后,我们还在经典机器学习模型基础上做一定改进,实验表明,改进后的模型预测效果进一步增强。

Price prediction is classic machine learning problem.Lots of research has been done in the past.The difference between this research and the previous ones is,they only use few flights and few months’data to research,which can’t express the complexity and periodicity of the flight business,and we have all airlines and long term data,this is a huge advantage.Furthermore,as flight ticket seller,we have deep un⁃derstanding of the flight price formation mechanism and the price changing trends,which help us make better use the data and create better model.We use Random Forest and XGBoost separately to build the prediction model,they both get good effects.At last,we optimize the model by stacking the two above,and get a boost.The experiments demonstrate that the optimized model can achieve better prediction per⁃formance.