Databases and Information Systems Commons^™

A Family Of Simple Non-Parametric Kernel Learning Algorithms From Pairwise Constraints, Jinfeng Zhuang, Ivor W. Tsang, Steven C. H. Hoi

Research Collection School Of Computing and Information Systems

Previous studies of Non-Parametric Kernel Learning (NPKL) usually formulate the learning task as a Semi-Definite Programming (SDP) problem that is often solved by some general purpose SDP solvers. However, for N data examples, the time complexity of NPKL using a standard interior-point SDP solver could be as high as O(N6.5), which prohibits NPKL methods applicable to real applications, even for data sets of moderate size. In this paper, we present a family of efficient NPKL algorithms, termed "SimpleNPKL", which can learn non-parametric kernels from a large set of pairwise constraints efficiently. In particular, we propose two efficient SimpleNPKL algorithms. One …

Confidence Weighted Mean Reversion Strategy For On-Line Portfolio Selection, Bin Li, Steven C. H. Hoi, Peilin Zhao, Vivek Gopalkrishnan

Research Collection School Of Computing and Information Systems

On-line portfolio selection has been attracting increasing attention from the data mining and machine learning communities. All existing on-line portfolio selection strategies focus on the first order information of a portfolio vector, though the second order information may also be beneficial to a strategy. Moreover, empirical evidences show that the stock price relatives may follow the mean reversion property, which has not been fully exploited by existing strategies. This article proposes a novel on-line portfolio selection strategy named ``Confidence Weighted Mean Reversion'' (CWMR). Inspired by the mean reversion principle in finance and confidence weighted online learning technique in machine learning, …

Two-Layer Multiple Kernel Learning, Jinfeng Zhuang, Ivor W. Tsang, Steven C. H. Hoi

Research Collection School Of Computing and Information Systems

Multiple Kernel Learning (MKL) aims to learn kernel machines for solving a real machine learning problem (e.g. classification) by exploring the combinations of multiple kernels. The traditional MKL approach is in general “shallow” in the sense that the target kernel is simply a linear (or convex) combination of some base kernels. In this paper, we investigate a framework of Multi-Layer Multiple Kernel Learning (MLMKL) that aims to learn “deep” kernel machines by exploring the combinations of multiple kernels in a multi-layer structure, which goes beyond the conventional MKL approach. Through a multiple layer mapping, the proposed MLMKL framework offers higher …

Learning Feature Dependencies For Noise Correction In Biomedical Prediction, Ghim-Eng Yap, Ah-Hwee Tan, Hwee Hwa Pang

Research Collection School Of Computing and Information Systems

The presence of noise or errors in the stated feature values of biomedical data can lead to incorrect prediction. We introduce a Bayesian Network-based Noise Correction framework named BN-NC. After data preprocessing, a Bayesian Network (BN) is learned to capture the feature dependencies. Using the BN to predict each feature in turn, BN-NC estimates a feature's error rate as the deviation between its predicted and stated values in the training data, and allocates the appropriate uncertainty to its subsequent findings during prediction. BN-NC automatically generates a probabilistic rule to explain BN prediction on the class variable using the feature values …

Predicting Item Adoption Using Social Correlation, Freddy Chong-Tat Chua, Hady W. Lauw, Ee Peng Lim

Research Collection School Of Computing and Information Systems

Users face a dazzling array of choices on the Web when it comes to choosing which product to buy, which video to watch, etc. The trend of social information processing means users increasingly rely not only on their own preferences, but also on friends when making various adoption decisions. In this paper, we investigate the effects of social correlation on users’ adoption of items. Given a user-user social graph and an item-user adoption graph, we seek to answer the following questions: 1) whether the items adopted by a user correlate to items adopted by her friends, and 2) how to …

Users Positions In Social Networks, Jasim Qazi

Master's Projects

Social networks are a new phase in human interaction: using
technology to connect people online, the social nehvorks of today have
become a central part of the lives of millions of people. People use social
networks for sharing various infonrmation with their friends and family. This
information can take the forn of text, video, images, sound etc. and it is what
forms the collection of dats in social networks.

As social networks gain popularity and as more and more people start
using social networks, it has become more important now to understand the
inner structures of social networks and understand …

Recipe Suggestion Tool, Sakuntala Padmapriya Gangaraju

Master's Projects

ABSTRACT
There is currently a great need for a tool to search cooking recipes based on ingredients. Current search engines do not provide this feature. Most of the recipe search results in current websites are not efficiently clustered based on relevance or categories resulting in a user getting lost in the huge search results presented.
Clustering in information retrieval is used for higher efficiency and better presentation of information to the user. Clustering puts similar documents in the same cluster. If a document is relevant to a query, then the documents in the same cluster are also relevant.
The goal …

Association Rule Mining -- Geometry And Parallel Computing Approach, Dongyi Jia

Master's Projects

Mining association rules is a very important aspect in data mining fields. The process to mine association rules not only take much time, but also take huge computing source. How to fast and efficiently find the large itemsets is a crucial point in the association rule algorithms. This paper will focus on two algorithms research and implementation in parallel computing environments. One is Bitmap Combination algorithm, the other is Bitmap FP-Growth algorithm. Compared to Apriori algorithm, both Bitmap Combination and Bitmap FP-Growth algorithms don’t need generate candidate items, avoids costly database scans. Both algorithms need to translate the original database …

Smart Search: A Firefox Add-On To Compute A Web Traffic Ranking, Vijaya Pamidi

Master's Projects

Search engines results are typically ordered according to some notion of importance of a web page as well as relevance of the content of a web page to a query. Web page importance is usually calculated based on some graph theoretic properties of the web. Another common technique to measure page importance is to make use of the traffic that goes to a particular web page as measured by a browser toolbar. Currently, there are some traffic ranking tools available like www.alexa.com, www.ranking.com, www.compete.com that give such analytic as to the number of users who visit a web site. Alexa …

Medical Analysis Question And Answering Application For Internet Enabled Mobile Devices, Loc Nguyen

Master's Projects

Mobile devices such as smart phones, the iPhone, and the iPad have become more popular in recent years. With access to the Internet through cellular or WIFI networks, these mobile devices can make use of the great source of information available on the Internet. Unlike a desktop or laptop computer, an Internet enabled mobile device is designed to be carried around and available to the owner almost instantly at any moment of the day. Despite having such great advantage and potential, searching for information with a mobile device remains a difficult task. Mobile device users have to juggle between different …

Identifying An Optimal Dining Plan System For The Entertainment Industry, Joseph Crimi

Doctoral Dissertations and Master's Theses

The Disney Dining Plan (DDP) is a pre-paid meal plan guests can purchase when they make their reservation at Walt Disney World (WDW). Under the current system, the information provided to guests explaining the program is unclear which leads to confusion for guests. For example, guests are not sure what food they can purchase using the DDP or at which dining locations they can use the DDP. Given these problems, the present study evaluated a new information system for the DDP. The independent variables in this study were symbol type, the symbols used in the current DDP and new symbols …

Efficient Topological Olap On Information Networks, Qiang Qu, Feida Zhu, Xifeng Yan, Jiawei Han, Philip Yu, Hongyan Li

Research Collection School Of Computing and Information Systems

We propose a framework for efficient OLAP on information networks with a focus on the most interesting kind, the topological OLAP (called “T-OLAP”), which incurs topological changes in the underlying networks. T-OLAP operations generate new networks from the original ones by rolling up a subset of nodes chosen by certain constraint criteria. The key challenge is to efficiently compute measures for the newly generated networks and handle user queries with varied constraints. Two effective computational techniques, T-Distributiveness and T-Monotonicity are proposed to achieve efficient query processing and cube materialization. We also provide a T-OLAP query processing framework into which these …

Ir-Tree: An Efficient Index For Geographic Document Search, Zhisheng Li, Ken C. K. Lee, Baihua Zheng, Wang-Chien Lee, Dik Lun Lee, Xufa Wang

Research Collection School Of Computing and Information Systems

Given a geographic query that is composed of query keywords and a location, a geographic search engine retrieves documents that are the most textually and spatially relevant to the query keywords and the location, respectively, and ranks the retrieved documents according to their joint textual and spatial relevances to the query. The lack of an efficient index that can simultaneously handle both the textual and spatial aspects of the documents makes existing geographic search engines inefficient in answering geographic queries. In this paper, we propose an efficient index, called IR-tree, that together with a top-k document search algorithm facilitates four …

Comparing Twitter And Traditional Media Using Topic Models, Wayne Xin Zhao, Jing Jiang, Jianshu Weng, Jing He, Ee Peng Lim, Hongfei Yan, Xiaoming Li

Research Collection School Of Computing and Information Systems

Twitter as a new form of social media can potentially contain much useful information, but content analysis on Twitter has not been well studied. In particular, it is not clear whether as an information source Twitter can be simply regarded as a faster news feed that covers mostly the same information as traditional news media. In This paper we empirically compare the content of Twitter with a traditional news medium, New York Times, using unsupervised topic modeling. We use a Twitter-LDA model to discover topics from a representative sample of the entire Twitter. We then use text mining techniques to …

Utility-Oriented K-Anonymization On Social Networks, Yazhe Wang, Long Xie, Baihua Zheng, Ken C. K. Lee

Research Collection School Of Computing and Information Systems

"Identity disclosure" problem on publishing social network data has gained intensive focus from academia. Existing k-anonymization algorithms on social network may result in nontrivial utility loss. The reason is that the number of the edges modified when anonymizing the social network is the only metric to evaluate utility loss, not considering the fact that different edge modifications have different impact on the network structure. To tackle this issue, we propose a novel utility-oriented social network anonymization scheme to achieve privacy protection with relatively low utility loss. First, a proper utility evaluation model is proposed. It focuses on the changes on …

War Fighting In Cyberspace: Evolving Force Presentation And Command And Control, M. Bodine Birdwell, Robert F. Mills

Faculty Publications

The Department of Defense (DOD) is endeavoring to define war fighting in the global cyberspace domain. Creation of US Cyber Command (USCYBERCOM), a subunified functional combatant command (FCC) under US Strategic Command (USSTRATCOM), is a huge step in integrating and coordinating the defense, protection, and operation of DOD networks; however, this step does not mean that USCYBERCOM will perform or manage all cyberspace functions. In fact the vast majority of cyberspace functions conducted by the services and combatant commands (COCOM), although vital for maintaining access to the domain in support of their operations, are not of an active war-fighting nature. …

Propeller: A Scalable Metadata Organization For A Versatile Searchable File System, Lei Xu, Hong Jiang, Xue Liu, Lei Tian, Yu Hua, Jian Hu

School of Computing: Technical Reports

The exponentially increasing amount of data in file systems has made it increasingly important for users, administrators and applications to be able to fast retrieve files using file-search services, instead of replying on the standard file system API to traverse the hierarchical namespaces.  The quality of the file-search services  is significantly affected by the file-indexing overhead, the file-search performance and the accuracy of search results.  Unfortunately, the existing file-search solutions either are so poorly scalable that their performance degrades unacceptably when the systems scale up, or incur so much crawling delays that they produce acceptably inaccurate results.  We believe that …

Identifying And Implementing The Underlying Operators For Nuclear Magnetic Resonance Based Metabolomics Data Analysis, Ashwin Manjunatha, Ajith H. Ranabahu, Paul E. Anderson, Amit P. Sheth

Kno.e.sis Publications

The science of metabolomics is a relatively young field that requires intensive signal processing and multivariate data analysis for interpretation of experimental results. The lack of integration and standardization for metabolomics compounded by the complexity of the experimental data has lead to a fragmented research community. While efforts have been undertaken to approach these problems, the efforts to develop a set of standards for reporting processing and analysis procedures has stalled.

In this paper, we propose a set of fundamental operators for nuclear magnetic resonance(NMR) based metabolomics. These operators are implementation independent, and can be used to easily and precisely …

Georgia Knowledge Repository: Potential For The State, Marlee Givens

Giving Undergraduate Research a Worldwide Voice: Institutional Repositories as Publishers

A presentation on the Georgia Knowledge Repository, a search engine of Georgia-based institutional repositories.

More information at http://www.library.gatech.edu/gkr/

Underfunded And Understaffed: A Ground-Level View Of Ir Creation In Difficult Times, John Davison

Giving Undergraduate Research a Worldwide Voice: Institutional Repositories as Publishers

A discussion of the creation of OhioLINK, the institutional repository system for the Ohio University system.

More information at http://drc.ohiolink.edu/

Enhance And "Mobilize" The Library Catalog With Cloud Services, Jolinda Thompson

Himmelfarb Library Faculty Publications

Explores new products and services that make it possible to enhance and mobilize traditional library catalogs.

Modeling Link Formation Behaviors In Dynamic Social Networks, Viet-An Nguyen, Cane Wing-Ki Leung, Ee Peng Lim

Research Collection School Of Computing and Information Systems

Online social networks are dynamic in nature. While links between users are seemingly formed and removed randomly, there exists some interested link formation behaviors demonstrated by users performing link creation and removal activities. Uncovering these behaviors not only allows us to gain deep insights of the users, but also pave the way to decipher how social links are formed. In this paper, we propose a general framework to define user link formation behaviors using well studied local link structures (i.e., triads and dyads) in a dynamic social network where links are formed at different timestamps. Depending on the role a …

An Approach To Nearest Neighboring Search For Multi-Dimensional Data, Yong Shi, Li Zhang, Lei Zhu

Faculty Articles

Finding nearest neighbors in large multi-dimensional data has always been one of the research interests in data mining field. In this paper, we present our continuous research on similarity search problems. Previously we have worked on exploring the meaning of K nearest neighbors from a new perspective in PanKNN [20]. It redefines the distances between data points and a given query point Q, efficiently and effectively selecting data points which are closest to Q. It can be applied in various data mining fields. A large amount of real data sets have irrelevant or obstacle information which greatly affects the effectiveness …

Fraud Detection In Online Consumer Reviews, Nan Hu, Ling Liu, Vallbh Sambamurthy

Research Collection School Of Computing and Information Systems

Increasingly, consumers depend on social information channels, such as user-posted online reviews, to make purchase decisions. These reviews are assumed to be unbiased reflections of other consumers' experiences with the products or services. While extensively assumed, the literature has not tested the existence or non-existence of review manipulation. By using data from Amazon and Barnes & Noble, our study investigates if vendors, publishers, and writers consistently manipulate online consumer reviews. We document the existence of online review manipulation and show that the manipulation strategy of firms seems to be a monotonically decreasing function of the product's true quality or the …

Evolution Of Developer Collaboration On The Jazz Platform: A Study Of A Large Scale Agile Project, Subhajit Datta, Renuka Sindhgatta, Bikram Sengupta

Research Collection School Of Computing and Information Systems

Collaboration is a key aspect of the agile philosophy of software development. As a software system matures over iterations, trends of developer collaboration can offer valuable insights into project dynamics. In this paper, we study evolution of developer collaboration for a large scale agile project on the Jazz platform. We construct networks of collaboration based on developer affiliations across comments on work items and file changes; and then compare parameters of such networks with established results from networks of scientific collaborations. The comparisons illuminate interesting facets of developer collaboration on the Jazz platform. Such perception helps deeper understanding of the …

Artificial Cognitive Memory - Changing From Density Driven To Functionality Driven, Luping Shi, Kaijun Yi, Kiruthika Ramanathan, Rong Zhao, Ning Ning, Ding Ding, Tow Chong Chong

Research Collection School Of Computing and Information Systems

Increasing density based on bit size reduction is currently a main driving force for the development of data storage technologies. However, it is expected that all of the current available storage technologies might approach their physical limits in around 15 to 20 years due to miniaturization. To further advance the storage technologies, it is required to explore a new development trend that is different from density driven. One possible direction is to derive insights from biological counterparts. Unlike physical memories that have a single function of data storage, human memory is versatile. It contributes to functions of data storage, information …

Fraud Detection In Online Consumer Reviews, Nan Hu, Ling Liu, Vallabh Sambamurthy

Research Collection School Of Computing and Information Systems

Increasingly, consumers depend on social information channels, such as user-posted online reviews, to make purchase decisions. These reviews are assumed to be unbiased reflections of other consumers' experiences with the products or services. While extensively assumed, the literature has not tested the existence or non-existence of review manipulation. By using data from Amazon and Barnes & Noble, our study investigates if vendors, publishers, and writers consistently manipulate online consumer reviews. We document the existence of online review manipulation and show that the manipulation strategy of firms seems to be a monotonically decreasing function of the product's true quality or the …

Distance Metric Learning From Uncertain Side Information For Automated Photo Tagging, Lei Wu, Steven C. H. Hoi, Rong Jin, Jianke Zhu, Nenghai Yu

Research Collection School Of Computing and Information Systems

Automated photo tagging is an important technique for many intelligent multimedia information systems, for example, smart photo management system and intelligent digital media library. To attack the challenge, several machine learning techniques have been developed and applied for automated photo tagging. For example, supervised learning techniques have been applied to automated photo tagging by training statistical classifiers from a collection of manually labeled examples. Although the existing approaches work well for small testbeds with relatively small number of annotation words, due to the long-standing challenge of object recognition, they often perform poorly in large-scale problems. Another limitation of the existing …

Mining Social Images With Distance Metric Learning For Automated Image Tagging, Pengcheng Wu, Steven C. H. Hoi, Peilin Zhao, Ying He

Research Collection School Of Computing and Information Systems

With the popularity of various social media applications, massive social images associated with high quality tags have been made available in many social media web sites nowadays. Mining social images on the web has become an emerging important research topic in web search and data mining. In this paper, we propose a machine learning framework for mining social images and investigate its application to automated image tagging. To effectively discover knowledge from social images that are often associated with multimodal contents (including visual images and textual tags), we propose a novel Unified Distance Metric Learning (UDML) scheme, which not only …

A Two-View Learning Approach For Image Tag Ranking, Jinfeng Zhuang, Steven C. H. Hoi

Research Collection School Of Computing and Information Systems

Tags of social images play a central role for text-based social image retrieval and browsing tasks. However, the original tags annotated by web users could be noisy, irrelevant, and often incomplete for describing the image contents, which may severely deteriorate the performance of text-based image retrieval models. In this paper, we aim to overcome the challenge of social tag ranking for a corpus of social images with rich user-generated tags by proposing a novel two-view learning approach. It can effectively exploit both textual and visual contents of social images to discover the complicated relationship between tags and images. Unlike the …