Published Paper Details:

Accurately predicting taxi ride durations is crucial for optimizing the efficiency of taxi dispatch systems in ride-hailing services. This research aims to develop and evaluate predictive models that estimate taxi ride durations before the start of the trip, leveraging historical ride data. Various machine learning techniques, including linear regression, K-Nearest Neighbors (KNN), Ridge regression, and Lasso regression, are explored using data from New York City taxi trips.The study aims to understand the key factors influencing ride durations and build models that can provide accurate estimates. Accurate predictions can improve dispatch decisions, reduce idle times for drivers, and decrease wait times for passengers, thus enhancing operational efficiency and service quality in the ride-hailing industry.We collect a comprehensive dataset of taxi trips, enriched with external data such as weather conditions and traffic information. The data undergoes preprocessing steps like cleaning, normalizing, and feature engineering. Key variables influencing ride durations include trip distance, pickup time, day of the week, weather conditions, and traffic patterns.The modeling approach involves developing predictive models, including a benchmark linear regression model, a KNN model, and regularized linear models such as Ridge and Lasso regression. Hyperparameter tuning optimizes model performance, and cross-validation ensures robustness. Performance is evaluated using metrics like Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared. The linear regression model serves as a baseline, capturing some variability in ride durations but limited in modeling non-linear interactions. The KNN model improves performance by considering data point similarity, while Ridge and Lasso regression models enhance accuracy by incorporating regularization terms. Lasso also performs feature selection by setting some coefficients to zero.This study highlights the importance of feature engineering and real-time data integration. Future research should focus on integrating real-time data sources, such as live traffic updates and weather conditions, to improve prediction accuracy further. Advanced machine learning techniques, such as ensemble methods and deep learning, should be explored to capture more complex relationships in the data.Ethical considerations and data privacy are critical in the use of predictive models in ride-hailing services. Ensuring passenger and driver privacy, and maintaining transparency and fairness in algorithmic decision-making, are essential for responsible use of data-driven models.