QUADRIVEN: A Framework for Qualitative Taxi Demand Prediction Based on Time-Variant Online Social Network Data...

Area of Science:

• Urban planning and transportation science
• Data mining and machine learning
• Environmental science and air quality management

Background:

• Road traffic pollution significantly impacts urban air quality.
• Efficient public transport, including taxis, is crucial for urban mobility.
• Existing taxi demand prediction models often rely on historical data, limiting interpretability and data availability.

Purpose of the Study:

• To introduce QUADRIVEN (QUalitative tAxi Demand pRediction based on tIme-Variant onlinE social Network data analysis), a novel approach for taxi demand prediction.
• To leverage human-generated data from online social networks for more interpretable taxi demand forecasting.
• To address the limitations of traditional regression-based taxi demand prediction methods.

Main Methods:

• Utilizing time-variant online social network data for taxi demand prediction.
• Developing a qualitative prediction framework based on categorical labels.
• Testing the QUADRIVEN approach with various machine learning models in a large urban setting.

Main Results:

• QUADRIVEN demonstrated promising results in predicting taxi demand.
• The approach achieved an F1 score exceeding 0.8 in a large urban area.
• Qualitative predictions provided semantically-enriched outputs, enhancing interpretability.

Conclusions:

• QUADRIVEN offers a viable alternative to traditional taxi demand prediction methods.
• Social network data provides a valuable resource for urban transport optimization.
• The qualitative approach enhances the understanding and application of taxi demand forecasts.