Mobility Analytics for Spatio-Temporal and Social Data

Communities, which are prevalent in attributed graphs (e.g., social networks and knowledge bases) can be used in emerging applications such as product advertisement and setting up of social events. Given a graph G and a vertex \(q \in G\), the community search (CS) query returns a subgraph of G that contains vertices related to q. In this article, we study CS over two common attributed graphs, where (1) vertices are associated with keywords; and (2) vertices are augmented with locations. For keyword-based attributed graphs, we investigate the keyword-based attributed community (or KAC) query, which returns a KAC for a query vertex. A KAC satisfies both structure cohesiveness (i.e., its vertices are tightly connected) and keyword cohesiveness (i.e., its vertices share common keywords). For spatial-based attributed graphs, we aim to find the spatial-aware community (or SAC), whose vertices are close structurally and spatially, for a query vertex in an online manner. To enable efficient KAC search and SAC search, we propose efficient query algorithms. We also perform experimental evaluation on large real datasets, and the results show that our methods achieve higher effectiveness than the state-of-the-art community retrieval algorithms. Moreover, our solutions are faster than baseline approaches. In addition, we develop the C-Explorer system to assist users in extracting, visualizing, and analyzing KACs.

The maritime Automatic Identification System (AIS) data is obtained from many different terrestrial and satellite sources. AIS data enables safety, security, environmental protection and the economic efficiency of the maritime sector. The quality of AIS receivers is not controlled in the same manner as AIS transmitters. This has led to a situation where AIS data is not as clean as it should/could be. Added to this is the lack of accuracy and standards in entering the voyage data by the mariners such as next port of call into the AIS equipment installed on vessels. By using analytics IMIS Global Limited has been able to process the AIS data stream to eliminate a large portion of the faulty data. This has allowed the resultant AIS data to be used for more accurate detailed analysis such as the long-term vessel track, port arrival events and port departure events. New data that is derived from processing AIS data has enhanced the information available to maritime authorities enabling a significant increase in safety, security, environmental protection and economic growth. The next generation of maritime data communications technology being based AIS. This is known as the VHF Data Exchange System (VDES) and this technology now enables further opportunities. The value from the large volumes of AIS data is extracted by visual, streaming, historical and prescriptive data analytics. The datAcron project is showing the way with regards to the processing and use of AIS and resultant trajectory data.

People’s attention tends to be drawn by important, or unique events, such as concerts, demonstrations, major football games, and others. Many individuals are even willing to travel long distances in order to attend events they regard as important. As a result, the everyday patterns that a person has, changes. This includes changes in the normal mobility patterns of this person, as well as changes in their social activities. In this work, we study these phenomena by analyzing the behavior of social media users. We investigate the activity and movement of users that either attend a unique event, or visit an important location, and contrast those to users that do not. Furthermore, based on the online activity of users that attend an event, we study the information that we can extract related to the mobility of these users. This information reveals some important characteristics that can be useful for a variety of location-based applications.

It is found that networks in real world divide naturally into communities or modules. Many community detection algorithms have been developed to uncover the community structure in networks. However, most of them focus on non-overlapping communities and the applicability of these work is limited when it comes to real world networks, which inherently are overlapping in most cases, e.g. Facebook and Weibo. In this paper, we propose an overlapping community detection algorithm based on edge representation learning. Firstly, we sample a series of edge sequences using random walks on graph, then a mapping function from edge to feature vectors is automatically learned in an unsupervised way. At last we employ the traditional clustering algorithms, e.g. K-means and its variants, on the learned representations to carry out community detection. To demonstrate the effectiveness of our proposed method, extensive experiments are conducted on a group of synthetic networks and two real world networks with ground truth. Experiment results show that our proposed method outperforms traditional algorithms in terms of evaluation metrics.

With the rapid growth of Online Social Networks (OSNs) and the information involved in them, research studies concerning OSNs, as well as the foundation of businesses, have become popular. Privacy on OSNs is typically protected by anonymisation methods. Current methods are not sufficient to ensure privacy and they impose restrictions on the network making it not suitable for research studies. This paper introduces an approach to find an optimal anonymous graph under user-defined metrics using Constraint Programming, a technique that provides well-tested and optimised engine for combinatorial problems. The approach finds a good trade-off between protection of sensitive data and quality of the information represented by the network.

The relations among heterogeneous data objects (e.g., image, tag, user and geographical point-of-interest (POI)) on interactive Online Social Media (OSM) play an important information source in describing complicated connections among Web entities (users and POIs) and items (images). Jointly predicting multiple relations instead of single relation completion in separate tasks facilitates sufficient knowledge sharing among heterogeneous relations and mitigate the information imbalance among different tasks. In this paper, we propose JEREMIE, a Joint SEmantic FeatuRe LEarning model via Multi-relational MatrIx ComplEtion, which jointly complements the semantic features of different entities from heterogeneous domains. Specifically, to perform appropriate information averaging, we first divide the social image collection into data blocks according to the affiliated user and POI information, where POIs are detected by mean shift from the GPS information. Then we develop a block-wise batch learning method which jointly learns the semantic features (e.g., image-tag, POI-tag and user-tag relations) by optimizing a transductive matrix completion framework with structure preservation and appropriate information averaging functionality. Experimental results on automatic image annotation, image-based user retrieval and image-based POI retrieval demonstrate that our approach achieves promising performance in various relation prediction tasks on six city-scale OSM datasets.

Unlike roads, shipping lanes are not carved in stone. Their size, boundaries and content vary over space and time, under the influence of trade and carrier patterns, but also infrastructure investments, climate change, political developments and other complex events. Today we only have a vague understanding of the specific routes vessels follow when travelling between ports, which is an essential metric for calculating any valid maritime statistics and indicators (e.g. trade indicators, emissions and others). Whilst in the past though, maritime surveillance had suffered from a lack of data, current tracking technology has transformed the problem into one of an overabundance of information, as huge amounts of vessel tracking data are slowly becoming available, mostly due to the Automatic Identification System (AIS). Due to the volume of this data, traditional data mining and machine learning approaches are challenged when called upon to decipher the complexity of these environments. In this work, our aim is to transform billions of records of spatiotemporal (AIS) data into information for understanding the patterns of global trade by adopting distributed processing approaches. We describe a four-step approach, which is based on the MapReduce paradigm, and demonstrate its validity in real world conditions.

Thanks to recent prevalence of location sensors, collecting massive spatiotemporal datasets containing moving object trajectories has become possible, providing an exceptional opportunity to derive interesting insights about behavior of the moving objects such as people, animals, and vehicles. In particular, mining patterns from co-movements of objects (such as movements by players of a sports team, joints of a person while walking, and cars in a transportation network) can lead to the discovery of interesting patterns (e.g., offense tactics of a sports team, gait signature of a person, and driving behaviors causing heavy traffic). Given a dataset of frequent co-movement patterns, various spatial and spatiotemporal queries can be posed to retrieve relevant patterns among all generated patterns from the pattern dataset. We term such queries, pattern queries. Co-movement patterns are often numerous due to combinatorial complexity of such patterns, and therefore, co-movement pattern datasets often grow very large in size, rendering naive execution of the pattern queries ineffective. In this paper, we propose the FCPIR framework, which offers a variety of index structures for efficient answering of various range pattern queries on massive co-movement pattern datasets, namely, spatial range pattern queries, temporal range (time-slice) pattern queries, and spatiotemporal range pattern queries.

Using evolving information within rapid mapping activities in the response phase of emergency situations poses a number of questions related to the quality of information being provided. In this paper, we focus on image extraction from social networks, in particular Twitter, in case of emergencies. In this case issues arise about the temporal and spatial location of images, which can be refined over time as information about the event is being collected and (automatically) analyzed. The paper describes a scenario for rapid mapping in an emergency event and how information quality can evolve over time. A model for managing and analyzing the evolving information is proposed to be used as a basis for analyzing the images quality for mapping purposes.

With the prevalence of mobile devices, millions of multimedia data represented as a combination of visual, aural and textual modalities, is produced every second. To facilitate better information retrieval on mobile devices, it becomes imperative to develop efficient models to retrieve heterogeneous content modalities using a specific query input, e.g., text-to-image or image-to-text retrieval. Unfortunately, previous works address the problem without considering the hardware constraints of the mobile devices. In this paper, we propose a novel method named Trigonal Partial Least Squares (TPLS) for the task of cross-modal retrieval on mobile devices. Specifically, TPLS works under the hardware constrains of mobile devices, i.e., limited memory size and no GPU acceleration. To take advantage of users’ tags for model training, we take the label information provided by the users as the third modality. Then, any two modalities of texts, images and labels are used to build a Kernel PLS model. As a result, TPLS is a joint model of three Kernel PLS models, and a constraint to narrow the distance between label spaces of images and texts is proposed. To efficiently learn the model, we use stochastic parallel gradient descent (SGD) to accelerate the learning speed with reduced memory consumption. To show the effectiveness of TPLS, the experiments are conducted on popular cross-modal retrieval benchmark datasets, and competitive results have been obtained.