PROJECT TITLE :

Mining the Most Influential k-Location Set From Massive Trajectories - 2017

ABSTRACT:

Mining the foremost influential location set finds $k$ locations, traversed by the utmost number of distinctive trajectories, in an exceedingly given spatial region. These influential locations are valuable for resource allocation applications, like choosing charging stations for electric cars and suggesting locations for putting billboards. This drawback is NP-arduous and typically involves an interactive mining processes involving a user's input, e.g., changing the spatial region and $k$, or removing some locations that are not eligible for an application per the domain data. Efficiency is the most important concern in conducting this human-in-the-loop mining. To this finish, we propose a complete mining framework, that includes an optimal method for the sunshine setting (i.e., tiny region and $k$) and an approximate methodology for the heavy setting (i.e., giant region and $k$). The optimal method leverages vertex grouping and best-1st pruning techniques to expedite the mining method. The approximate method can give the performance guarantee by utilizing the greedy heuristic, and it is comprised of economical updating strategy, index partition and workload-based optimization techniques. We evaluate the potency and effectiveness of our methods primarily based on two taxi datasets from China, and one check-in dataset from New York.