Share-a-Cab: Scalable Clustering Taxi Group Ride Stand From Huge Geolocation Data

Abstract

Taxi group ride service (TGRS) is one potentially successful way to make traditional services competitive as emerging app-based taxi services, simply through grouping similar taxi rides without significant budget increases, generating one unique pick-up point and one unique drop-off point, thus serving multiple passengers in one single trip. In this study, we mainly develop a scalable method for citywide TGRS stand deployment driven by huge traditional taxicab trips. First, a spatial temporal clustering method is proposed to explore trip clusters that present potential group rides. Second, the agglomerative clustering method is applied to merge trip clusters at both spatial and temporal scale, which will yield potential taxi stand location and schedule. Based on the one-month taxi trips in New York City, the proposed approach can fast process the huge dataset and identify more than 60 stands with four schedules. The study contributes towards efficient methods for developing TGRS in large-scale taxi systems.

Keywords

• Data mining
• geolocation data
• scalable stand deployment
• spatial-temporal clustering
• taxi group ride services