Evaluation and Optimization of Predictive Models for Taxi Service Metrics in New York City Using Integrated Traffic and Weather Data

This study analyzes data from New York City's yellow taxi service, integrating both taxi order and weather data, to predict the tipping ratio, congestion level, and order quantity using several modeling approaches: Linear Regression, Ridge Regression, Decision Tree, and Random Forest models. By optimizing the Decision Tree and Random Forest algorithms and conducting a horizontal comparison of the mean squared error (MSE) between training and testing datasets for each model, the research identified the most effective model for addressing different predictive challenges. The regression coefficients derived from the least squares method were examined to ascertain the magnitude and direction of various influencing factors on the prediction targets, thereby providing insights of practical significance. The findings of this research not only offer valuable guidance for urban transportation policy development but also contribute innovative perspectives on data feature engineering and model optimization. These insights hold substantial practical relevance for enhancing urban transportation management.