Riding the AI Wave: ARIMA-LSTM Hybrid and XGBoost Race to Perfect Fare Prediction

The rate of urbanization worldwide, with particular emphasis on several megacities, is rapidly growing and places a strong burden on transport infrastructure for the needs of the residence, commuters, and tourists. The taxi industry remains an integral part of the network that provides critical on-demand services complementing other modes of public transport. Taxi demand forecasting is thus critically important to optimize these services and thus guarantee efficiency and reliability. Advanced machine learning models, namely XGBoost, ARIMA, and a hybrid ARIMA-LSTM model, were used in the current work for taxi demand estimation. Evaluation of the predictive accuracy in combination with practical usability of the considered models may become the most efficient means for developing data-driven transport systems in real-world conditions.