Fei Yi

Moving Destination Prediction Using Sparse Dataset: A Mobility Gradient Descent Approach

Moving destination prediction offers an important category of location-based applications and provides essential intelligence to business and governments. In existing studies, a common approach to destination prediction is to match the given query trajectory with massive recorded trajectories by similarity calculation. Unfortunately, due to privacy concerns, budget constraints, and many other factors, in most circumstances, we can only obtain a sparse trajectory dataset. In sparse dataset, the available moving trajectories are far from enough to cover all possible query trajectories; thus the predictability of the matching-based approach will decrease remarkably. Toward destination prediction with sparse dataset, instead of searching similar trajectories over the sparse records, we alternatively examine the changes of distances from sampling locations to final destination on query trajectory. The underlying idea is intuitive: It is directly motivated by travel purpose, people always get closer to the final destination during the movement. By borrowing the conception of gradient descent in optimization theory, we propose a novel moving destination prediction approach, namely MGDPre. Building upon the mobility gradient descent, MGDPre only investigates the behavior characteristics of query trajectory itself without matching historical trajectories, and thus is applicable for sparse dataset. We evaluate our approach based on extensive experiments, using GPS trajectories generated by a sample of taxis over a 10-day period in Shenzhen city, China. The results demonstrate that the effectiveness, efficiency, and scalability of our approach outperform state-of-the-art baseline methods.

Group mobility classification and structure recognition using mobile devices

Monitoring group mobility and structure is crucial for public safety management and emergency evacuation. In this paper, we propose a fine-grained mobility classification and structure recognition approach for social groups based on hybrid sensing using mobile devices. First, we present a method which classifies group mobility into four levels, including stationary, strolling, walking and running. Second, by combining mobile sensing and Wi-Fi signals, a novel relative position relationship estimation algorithm is developed to understand moving group structures of different shapes. We have conducted real-life experiments in which eight volunteers form two to three small groups moving in a teaching building with different speed and structures. Experimental results show that our approach achieves an accuracy of 99.5% in mobility classification and about 80% in group structure recognition.

Cyber-physical-social collaborative sensing: from single space to cross-space

The development of wireless sensor networking, social networking, and wearable sensing techniques has advanced the boundaries of research on understanding social dynamics. Collaborative sensing, which utilizes diversity sensing and computing abilities across different entities, has become a popular sensing and computing paradigm. In this paper, we first review the history of research in collaborative sensing, which mainly refers to single space collaborative sensing that consists of physical, cyber, and social collaborative sensing. Afterward, we extend this concept into cross-space collaborative sensing and propose a general reference framework to demonstrate the distinct mechanism of cross-space collaborative sensing. We also review early works in cross-space collaborative sensing, and study the detail mechanism based on one typical research work. Finally, although cross-space collaborative sensing is a promising research area, it is still in its infancy. Thus, we identify some key research challenges with potential technical details at the end of this paper.

Toward estimating user-social event distance: mobility, content, and social relationship

On-site user w.r.t social events are valuable, from whom, government/police could obtain meaningful information which contributes to understand the progress of the event or investigate suspects when the event is associated with crime or terrorist. However, due to the high uncertainty of human mobility patterns, it is hard to identify on-site users while social event happens. In this paper, we propose a F>used fEature Gaussian prOcess Rgression (FEGOR) model, which employs three features from online social networks: mobility influence, content similarity, and social relationship to estimate the distance between user and social event, based on which, we could accomplish the problem of identifying the on-site users. Experiment results on a real-world Twitter dataset demonstrate our method outperforms state-of-the-art methods.

Mobile crowd sensing task optimal allocation: a mobility pattern matching perspective

With the proliferation of sensor-equipped portable mobile devices, Mobile CrowdSensing (MCS) using smart devices provides unprecedented opportunities for collecting enormous surrounding data. In MCS applications, a crucial issue is how to recruit appropriate participants from a pool of available users to accomplish released tasks, satisfying both resource efficiency and sensing quality. In order to meet these two optimization goals simultaneously, in this paper, we present a novel MCS task allocation framework by aligning existing task sequence with users’ moving regularity as much as possible. Based on the process of mobility repetitive pattern discovery, the original task allocation problem is converted into a pattern matching issue, and the involved optimization goals are transformed into pattern matching length and support degree indicators. To determine a trade-off between these two competitive metrics, we propose greedy-based optimal assignment scheme search approaches, namely MLP, MDP, IU1 and IU2 algorithm, with respect to matching length-preferred, support degree-preferred and integrated utility, respectively. Comprehensive experiments on realworld open data set and synthetic data set clearly validate the effectiveness of our proposed framework on MCS task optimal allocation.

Fine-grained Emotion Role Detection Based on Retweet Information

User behaviors in online social networks convey not only literal information but also one’s emotional attitudes towards the information. To compute this attitude, we define the concept of emotion role as the concentrated reflection of a user’s online emotional characteristics. Emotion role detection aims to better understand the structure and sentiments of online social networks and support further analysis, e.g., revealing public opinions, providing personalized recommendations, and detecting influential users. In this article, we first introduce the definition of a fine-grained emotion role, which consists of two dimensions: emotion orientation (i.e., positive, negative, and neutral) and emotion influence (i.e., leader and follower). We then propose a Multi-dimensional Emotion Role Mining model (MERM) to determine a user’s emotion role in online social networks. Specifically, we tend to identify emotion roles by combining a set of features that reflect a user’s online emotional status, including degree of emotional characteristics, accumulated emotion preference, structural factor, temporal factor, and emotion change factor. Experiment results on a real-life micro-blog reposting dataset show that the classification accuracy of the proposed model can achieve up to 90.1%.

Discovery of booming and decaying point-of-interest with human mobility data

By observing the Point-Of-Interests (POIs) in cities over time, we can find that some of them disappear, new ones emerge and most of them just keep alive. In this paper, we name such evolutionary process of POI as POI lifetime and the stages of POI lifetime as POI lifetime status. Specifically, when we find a POI appears/disappears by observing the online maps over time, we may say that a POI is in booming/decaying lifetime status. How to discover the booming/decaying POIs is a challenging but valuable task. To address this problem, we propose a framework leveraging the human mobility data to discover the booming and decaying POIs in urban areas. We conduct the experiments on real-world data sets. The results demonstrate that the ensemble classifier can discover the booming/decaying POIs effectively, and the human mobility factors extracted are indicative for discovering the POI lifetime status.

Participant Selection for Information Diffusion Based on Topic and Emotion Preference Learning

The rapid development of social networks has woven themselves into peoples daily life and become indispensable platforms with superior commercial and scientific values. Both companies and governments have discovered the potential effectiveness of employing social network users for information diffusion. Rather than only selecting a group of users who are interested in target topic, it is more beneficial to choose users with desired emotion preference to help diffuse information under certain emotional expectation. In this paper, we propose an emotional participant selection system that not only considers users topic preference, but also more importantly takes users emotional influence into account. Specifically, a dynamic forgetting mechanism is applied to learn users topic preference, and independent cascade model is leveraged to construct emotional influence. Combining these two features, we develop an algorithm that can accomplish the task for emotional participant selection. Experimental results on a real-world data set validate the effectiveness of our proposed method.

Investigating collaboration evolution in UbiComp research

During the past 17 years, UbiComp has grown into one of the most influential conferences. To further figure out how the contributors of UbiComp collaborate with each other, we first construct a collaboration network based on 765 UbiComp papers published since 1999. Afterwards, to examine the patterns of collaborations, a PCA-based time series segmentation scheme is proposed, which suggests 2009 as a break point. By applying comparative analysis, we then investigate the collaboration evolution in different periods and propose a new method to quantify the actual effect of each paper. Experimental results reveal the existence of four types of collaboration patterns and proved that, as UbiComp getting more collaborative, long-term collaboration is more inclined to facilitate high quality achievements.

SmartSwim: An Infrastructure-Free Swimmer Localization System Based on Smartphone Sensors

Many works have focused their attention on the sports activity monitoring and recognition using inherit sensors on the smartphone. However, distinct from many on-the-ground activities, swimming is not only hard to monitor but also dangerous in the water. Knowing the position of a swimmer is crucial which can help a lot in rescuing people. In this paper, we propose a system called SmartSwim employing smartphone as a sensor for swimming tracking and localization. In detail, we first present a sensor based swimming status classification and moving length estimation. A swimmer locating algorithm is then proposed drawing on the experience of pedestrian dead reckoning PDR concept. We implemented the system on commercial smartphones and designed two prototype applications named WeSwim and SafeSwim. Experimental results showed the accuracy of swimming status classification reaches more than 99i¾ź% and the Error Rate value for length estimation is lower than 7i¾ź% overall.