Discovering Geographical Features for Location-Based Services

In applications such as location-based services, development planning and area marketing, the knowledge of frequent service requests that are always issued together is useful for decision and policy making. However, knowing merely the frequently co-located service requests may not suffice. We observe that often times, these co-located service requests are influenced by surrounding geographical features. By incorporating geographical features with the co-located service requests, we discover a new class of patterns called geographical-based NRS (Neighbouring service Request Sets), which is found to reveal more information compared to co-located service requests. We design two algorithms, namely TwoPhaseGSS and AprioriGSS, to discover this new class of patterns. Experiment results demonstrate the efficiency and the scalability of these algorithms.