STS: Complex Spatio-Temporal Sequence Mining in Flickr

Abstract

Nowadays, due to the increasing user requirements of efficient and personalized services, a perfect travel plan is urgently needed. In this paper we propose a novel complex spatio-temporal sequence (STS) mining in Flickr, which retrieves the optimal STS in terms of distance, weight, visiting time, opening hour, scene features, etc.. For example, when a traveler arrives at a city, the system endow every scene with a weight automatically according to scene features and user’s profiles. Then several interesting scenes (e.g., o ₁,o ₂,o ₃,o ₄,o ₅,o ₆) with larger weights (e.g., w ₁,w ₂,w ₃,w ₄,w ₅,w ₆) will be chosen. The goal of our work is to provide the traveler with the optimal STS, which passes through as many chosen scenes as possible with the maximum weight and the minimum distance within his travel time (e.g., one day). The difficulty of mining STS lies in the consideration of the weight of each scene, and its difference for different users, as well as the travel time limitation. In this paper, we provide two approximate algorithms: a local optimization algorithm and a global optimization algorithm. Finally, we give an experimental evaluation of the proposed algorithms using real datasets in Flickr.

This research was partially supported by the grants from the Natural Science Foundation of China (No.60833005, 61070055, 61003205); the National High-Tech Research and Development Plan of China (No.2009AA011904).