Optimizing Urban Mobility: A Comparative Analysis of Taxi Demand Prediction Models

Urban mobility optimization is crucial in managing transportation systems efficiently.This study addresses a broad research area of urban mobility by focusing on taxi demand prediction, a key component of the transportation ecosystem.The specific problem addressed in this research is the need for accurate and efficient taxi demand prediction, especially in large, dynamic urban environments.Existing solutions, including basic time series approaches like simple moving averages and exponential weighted moving averages, while valuable, have limitations in handling the intricacies of urban taxi demand patterns.In this study, we employed a combination of data preprocessing techniques, advanced regression models, and Fourier features to predict taxi demand in dynamic urban environments.The data preprocessing techniques included data cleaning, normalization, and feature engineering.The advanced regression models used in this study were Random Forest and XGBoost, which were trained and tested using NYC taxi datasets.The Fourier features were used to capture the periodicity of the taxi demand patterns.These models are demonstrated to outperform standard solutions, effectively achieving the targeted mean absolute percentage error (MAPE) of less than 12%.Evaluation of the solution revealed its effectiveness in reducing the prediction error by more than 1%, thus highlighting the positive results of this research.