Predictive spatio-temporal queries

Abstract

Predictive queries over spatio-temporal data proved to be vital in many location-based services including traffic management, ride sharing, and advertising. In the last few years, one of the most exciting work on spatio-temporal data management is about predictive queries. In this paper, we review the current research trends and present their related applications in the field of predictive spatiotemporal queries processing. Then, we discuss some basic challenges arising from new opportunities and open problems. The goal of this paper is to catch the interesting areas and future work under the umbrella of predictive queries over spatio-temporal data.

Predictive tree: An efficient index for predictive queries on road networks

“…However, existing techniques suffer from both functional limitations and performance deficiencies. From a functional perspective, they suffer from one or more of the following limitations: (1) They consider an Euclidean space [9], [28], [25], [30] where objects can move freely in a two dimensional space. Yet, practical predictive location-based services target moving objects on road networks as described by the motivating applications earlier in this section.…”

“…Yet, practical predictive location-based services target moving objects on road networks as described by the motivating applications earlier in this section. (2) Many techniques utilize prediction models that must be trained using a massive a mount of objects' historical trajectories in order to produce accurate predictions [2], [9], [13], [15], [24], [28]. However, practical scenarios and industrial experience reveal that such historical data is not easily obtainable for many reasons, either due to users' privacy and data confidentiality on one side or due to the unavailability of historical data in rural areas on the other side.…”

“…However, practical scenarios and industrial experience reveal that such historical data is not easily obtainable for many reasons, either due to users' privacy and data confidentiality on one side or due to the unavailability of historical data in rural areas on the other side. (3) Most of the previous solutions were designed to support a specific query type only, e.g., [13], [25], [30] support predictive range query, [3], [22], [30] support predictive KNN query, and [9], [24] support predictive aggregate query.…”

“…(3) Aggregate queries, i.e., [2], [9], [24]. A predictive aggregate query has a query region R and a future time window T , and asks about the number of objects N expected to be inside R within the time window T .…”

“…The approximate query processing approach in [24] employs an adaptive multidimensional histogram (AMH), a historical synopsis, and a stochastic method to provide an approximate answer for aggregate spatio-temporal queries about the future in addition to the past and the present. A generic query processing framework named Panda [9] is introduced to evaluate predictive queries for moving objects on Euclidean space.…”