Adaptive reinforcement learning of spatial interaction based on taxi trajectory data Available to Purchase

A novel approach for analysing spatial interaction characteristics and land use using taxi trajectory data and urban geographic data is introduced. An adaptive reinforcement learning model, based on non-linear theory, is proposed to improve the accuracy and adaptability of spatial interaction predictions. By dividing an urban area into smaller units, a spatial interaction matrix is constructed that captures push–pull force characteristics and distance features between origins and destinations. The innovative aspect of the model lies in its ability to integrate multiple weak spatial interaction learners to form a strong learner, thus significantly outperforming traditional models based on gravity theory in terms of prediction performance (higher R², lower mean absolute error and lower root mean squared error). The findings of this work reveal the importance of adjacent flows in predicting spatial interaction patterns and show that travel distance in public transportation is the most significant factor in describing the difficulty of completing spatial interactions. The push force from origins was found to have the highest relative importance, followed by the pull force from destinations and adjacent flows. The results of this study provide valuable insights for traffic and urban planning.