An Adaptive and Integrated Prediction of Taxi Trajectories using Regression Model

Abstract

Taxi trajectory prediction can help many advanced location-based services, such as advertisements for passengers. Previous approaches, which consider trajectories as one-dimensional patterns and process them in single scales, does not consider two-dimensional patterns of trajectories in various spatial scales. In road network analysis, one of the most main problems is finding important crossroads, which can be useful in transportation planning. However, none of the previous approaches identifies the problem of finding network-wide important crossroads in the actual road network. This paper proposes a strategic decision based on data analysis and interpretation, based approach to find important crossroads. This method is to model the road trip network focusing on real time travel demands using a tripartite graph, instead of completely analyzing on the topology of road network. A novel ranking algorithm that propagates scores along a tripartite graph is proposed to find the importance scores of crossroads accurately.