- banner1
Close Me!

Mining the world's photo

Yi Yang YANG

In recent years, users are used to share their photos to the others via online service, such as Flickr and Facebook. The photos uploaded by the users provide not only visual features but also rich temporal and spatial information. Such a large amount of user-contributed data can be used for deferent mining tasks. Extracting POIs from online photos is one of the examples. Previous work only utilize simple clustering approaches, such as K-means, DBSCAN, and mean shift. However, none of these approaches can extract POIs precisely and concisely. In this work, we aim at development of high quality POIs extraction using the spatial and temporal information of the online photos.

  • Mining the world's photo

    Mining the World's Photo

    Yi Yang YANG

    In recent years, users are used to share their photos to the others via online service, such as Flickr and Facebook. The photos uploaded by the users provide not only visual features but also rich temporal and spatial information. Such a large amount of user-contributed data can be used for deferent mining tasks. Extracting POIs from online photos is one of the examples. Previous work only utilize simple clustering approaches, such as K-means, DBSCAN, and mean shift. However, none of these approaches can extract POIs precisely and concisely. In this work, we aim at development of high quality POIs extraction using the spatial and temporal information of the online photos.

  • Business document interoperability&inference

    Business Document Interoperability and Inference

    Guang Yi XIAO

    In e-marketplace, enterprise users communicate with each other using diverse types of documents. However, these documents are written by heterogeneous templates and in different languages. Therefore, the users from different enterprises are hard to understand the documents in different systems. This leads to un-interoperable. We propose some novel approaches that can make business document interoperable in e-marketplace and data exchange.

  • abstract for trust-rank

    Trust-Rank

    Hai Tao Zou

    The need of online recommender systems is fast advancing due to the explosive growth of World Wide Web. Such systems filter out some items in which users are not interested to achieve better navigation experiences. Both user-based and item-based algorithms are studied extensively in previous work. However, none of them can handle the cold-start users who have a limited number of records in the system. In this work, we propose a novel algorithm which utilizes Trust-Rank to increase the performance for cold-start users.

  • Evolution

    Evaluating Review Quality using Social Context

    Wei Shu HU

    The quality of a online review is strongly related to the quality of the reviewer. In our preliminary result, we observe that the quality of reviewers can be analyzed by their social network connection. It is very interesting to study how the reviewer are influenced by their friends in the social networks and the relations to the quality of reviews

  • Hidden database

    Data Analysis in Hidden Database

    Hui Yan

    Database behind the web interface is called hidden database. As hidden database containing good quality of data, we would like to know the properties of this invisible data, such as the aggregate information of hidden database (e.g., size, sum, etc.). In this work, we consider a sampling problem in hidden transactional databases.

  • Stime series correlation

    Towards Online Shortest Paths Computation

    Hongjun Zhao

    With the rapid development of wirless network and related mobile devices, location-based services are becoming more popular nowadays, such as Google map, GPS and so on. Shortest Path problem (SP) is the most prevalent one among such services. However, there is no solution to deal with SP problem with live traffic information on huge road networks. In this work, we aim at solving the scalability and dynamic update problems of Online Shortest Path problem(OSP).

  • Stime series correlation

    Time Series Correlation Analysis

    Yuhong Li

    Time series data are often too large in their raw format and hard to be used for decision making. There are quite a lot of queries proposed for processing time series data, i.e., correlation. The correlation similarity measure between two time series is more robust against data that isn't normalized than over the Euclidean Distance. And recent years has seen a lot of work study the pair-wise correlation bettween the time series. More interesting problem is to find the longest subsequence having the correlation over a threshold between the query and the time series datasets. In this work, we study a novel algorithm that can identify the longest correlated subsequence efficiently for large scale time series data.

News & Highlights

A new paper gets accept in SIGIR 2011
A new paper gets accept in SIGIR 2011 (10 July, 2011)

Yiyang Yang, Zhiguo Gong and Leong Hou U have had a paper accepted to the 34th Annual ACM SIGIR Conference (SIGIR'2011). The paper is titled: Identifying Points of Interest by Self-Tuning Clustering.

Prof. Qiang Yang visited DEGroup
Prof. Qiang Yang visited DEGroup (15 May, 2011)

Prof. Qiang Yang from the Department of Computer Science and Engineering, Hong Kong University of Science and Technology visited DEGroup at 13 May 2011 and delivered a talk titled "Transfer Learning in Text, Multimedia and Networks". Prof. Qiang Yang is an IEEE Fellow. His research interests are data mining and artificial intelligence(automated planning, machine learning and activity recognition).

Dr. Man Lung YIU and Dr. Eric LO visited DEGroup
Dr. Man Lung YIU and Dr. Eric LO visited DEGroup (7 Jan, 2011)

Dr. Man Lung YIU and Dr. Eric LO from the Hong Kong Polytechnic University visited DEGroup and gave the seminars to the group members. Both of them are They are productive scholars in Databases and Data Mining communities.After the seminars Dr. Man Lung YIU and Dr. Eric LO joined the group meeting and gave valuable comments and suggestions on our work.

Recent Publications

Yiyang Yang, Zhiguo Gong and Leong Hou U, "Identifying Points of Interest by Self-Tuning Clustering".
SIGIR'2011,10 July 2011