Search

Fast-developing single-cell multimodal omics (scMulti-omics) technologies enable the measurement of multiple modalities, such as DNA methylation, chromatin accessibility, RNA expression, protein abundance, gene perturbation, and spatial information, from the same cell. scMulti-omics can comprehensively explore and identify cell characteristics, while also presenting challenges to the development of computational methods and tools for integrative analyses. Here, we review these integrative methods and summarize the existing tools for studying a variety of scMulti-omics data. The various functionalities and practical challenges in using the available tools in the public domain are explored through several case studies. Finally, we identify remaining challenges and future trends in scMulti-omics modeling and analyses.

Several recent studies of ERP system interfaces have confirmed that their poor usability hinders worker productivity, despite the huge investments companies make in user training and support. Usability challenges arise from the complexity of ERP systems, which are designed as a universal tool for a plethora of organizational practices and contexts. Learning to operate within an excessively vast terrain of ERP task pages and parameters is a significant challenge for most ERP users. Our proposed solution relies on the system itself to share task and process information in order to guide users through learning and performing their business tasks with the system. This perspective arises from employing the human–computer collaboration approach to the design of user interfaces, which we apply as a guiding framework for our research. In this paper, we present two interface components for providing ERP system users with task and process guidance: Automated Playback and Interactive Process Visualization. The novelty of our approach comes from using the history of past interactions to dynamically compose animated demonstrations of task interfaces and to provide an interactive graphical map of the current process being worked on by the user.

Motivation The biclustering of large-scale gene expression data holds promising potential for detecting condition-specific functional gene modules (i.e. biclusters). However, existing methods do not adequately address a comprehensive detection of all significant bicluster structures and have limited power when applied to expression data generated by RNA-Sequencing (RNA-Seq), especially single-cell RNA-Seq (scRNA-Seq) data, where massive zero and low expression values are observed. Results We present a new biclustering algorithm, QUalitative BIClustering algorithm Version 2 (QUBIC2), which is empowered by: (i) a novel left-truncated mixture of Gaussian model for an accurate assessment of multimodality in zero-enriched expression data, (ii) a fast and efficient dropouts-saving expansion strategy for functional gene modules optimization using information divergency and (iii) a rigorous statistical test for the significance of all the identified biclusters in any organism, including those without substantial functional annotations. QUBIC2 demonstrated considerably improved performance in detecting biclusters compared to other five widely used algorithms on various benchmark datasets from E.coli, Human and simulated data. QUBIC2 also showcased robust and superior performance on gene expression data generated by microarray, bulk RNA-Seq and scRNA-Seq. Availability and implementation The source code of QUBIC2 is freely available at https://github.com/OSU-BMBL/QUBIC2. Supplementary information Supplementary data are available at Bioinformatics online.

Prior research has found the beauty premium effect in peer-to-peer (P2P) lending markets. However, it remains unknown whether and how lenders rely on borrowers’ facial characteristics to infer their financial credibility. In this research, we propose a set of facial features that can be automatically extracted from images using face recognition techniques. Our analysis of a large sample collected from a Chinese P2P lending platform reveals that borrowers with lower socioeconomic status are more likely to include face images in their account profiles. However, these images do not necessarily help attract more investors. The in-depth analysis of the facial features shows that when presented with face images of borrowers, lenders tend to associate the appearance of borrowers with their financial credibility and generally prefer “good-looking” borrowers, who have a round face, a longer philtrum, and a higher nose. However, these facial features cannot be used to predict loan repayment performance.

Taking the data deluge from social media, in its many forms, and transforming it into useful information, or knowledge, is the essence of this minitrack. The papers at HICSS in 2015 remind our attendees and readers of the many real-world applications of data analytics and data mining for social media. Last year, for example, the papers showed how to aggregate data to discover the location of tweeters as well as how source code and prose can be combined to yield more valuable Stack Overflow posts. In 2013, our papers explored conducting a social outreach campaign to help prevent child abuse, and also analyzing travel patterns based on geo-tagged photographs. This minitrack begins with “A meta-analysis of theories and topics in social media research,” by Van Osch and Coursaris. This study provides an overview of the identity and intellectual core of social media research with respect to its prevailing theory-related practices. As such, this study gives us an important opportunity to pause and reflect on what has been accomplished by social media scholars to date. The findings provide a benchmark for tracking the state of the social media domain, while focusing our attention on topics and theories requiring further inquiry. The second paper explores if Twitter trends can predict election results based on evidence from the 2014 general election in India. Khatua, Khatua, Ghosh and Chaki show that that tweet volume and sentiment analysis can predict election results. This author team also finds that sentiment scoring can predict changes in voter share, even during an election. Finally, this study reaffirms the need for contextual understanding in forecasting election results. The third paper, “TreeQueST: A treemap-based query sandbox for microdocument retrieval,” is by Thom and Ertl. This University of Stuttgart team shows that the rise of online microdocument platforms, such as Twitter, has brought new relevance to techniques for finding and understanding information about recent events. The authors present and evaluate an approach for doing just this. Their novel approach builds on hierarchical topic clustering combined with a treemap-based visualization. “Viewer engagement in movie trailers and box office revenue,” by Oh, Ahn and Baek examines the antecedents of sharing movie trailers. Furthermore, this study assesses the impact of the consumption of and commentary on movie trailers on revenue. This work also highlights the practical challenges in attributing changes in box office revenue to consumer engagement on social media. The fifth paper, “Identifying uptake, sessions, and key actors in a socio-technical network,” is by Suthers and Dwyer. These authors address two challenges arising from the nature of participation in socio-technical networks. First, since learning and knowledge production takes place through both individual and collective agency, it is necessary to understand aggregate phenomenon (e.g., “ties”, “roles”, and “communities”) as both produced by and providing the setting for interactional events. Second, participant interaction is distributed across media, places and time, often resulting in separate traces of interaction that fragment their unitary experience. “Measuring NBA players’ mood by mining athlete-generated content,” by Xu and Yu, proposes a framework based on athlete generated content and assesses the relationship between an athlete’s mood and their individual performance. At the core of this framework is the use of Twitter as a source of intelligence and the choice of sentiment analysis to mine players’ moods from their tweets. This research also demonstrates the utility and limitations of mining user-generated unstructured text to predict outcomes such as individual performance. Our final paper is “Estimating centrality statistics for complete and sampled networks: Some approaches and complications” by Lee and Pfeffer. This Carnegie Mellon author team presents some approximation techniques exploiting any tractable relationship between the measures and network characteristics such as size and density. They find there exist distinct functional relationships between network statistics of complex “slow” measures and “fast” measures, such as the linkage between betweenness centrality and network density. Furthermore, they track how these relationships scale with network size. 2015 48th Hawaii International Conference on System Sciences

This exploratory study is intended to address the problem of fraudulent loan requests on peer-to-peer (P2P) platforms. We propose a set of features that capture the behavioral characteristics (e.g., learning, past performance, social networking, and herding manipulation) of malevolent borrowers, who intentionally create loan requests to acquire funds from lenders but default later on. We found that using the widely adopted classification methods such as Random Forest and Support Vector Machines, the proposed feature set outperform the baseline feature set in helping detect fraudulent loan requests. Although the performance (e.g., Recall or Sensitivity) is still not up to its optimum, this study demonstrates that by analyzing the transaction records of confirmed malevolent borrowers, it is possible to capture some useful behavioral patterns for fraud detection. Such features and methods would possibly help lenders identify loan request frauds and avoid financial losses.

Knowing the structure of criminal and terrorist networks could provide the technical insight needed to disrupt their activities.

As millions of people are diagnosed with diabetes every year, the demand for information about diabetes continues to increase. China is one of the countries with a large population of diabetes patients. Many Chinese health websites provide diabetes related news and articles. However, because most of the online articles are uncategorized or lack a clear topic and theme, users often cannot find their topics of interest effectively and efficiently. The problem of health topic identification and categorization on Chinese websites cannot be easily addressed by applying existing approaches and methods, which have been used for English documents, in a straightforward manner. To address this problem and meet users’ demand for diabetes related information needs, we propose a deep learning based framework to identify and categorize topics related to diabetes in online Chinese articles. Our experiments using datasets with over 19,000 online articles showed that the framework achieved a higher effectiveness and accuracy in categorizing diabetes related topics than most of the state-of-the-art benchmark approaches.

Chinese chronic disease entity extraction aims to extract health related entities from online questions and answers QA. Our research tackles challenges in Chinese chronic disease entity extraction from three aspects: Chinese health lexicons construction, feature development, and equivalence conjunctions tagging. We construct large scale Chinese health lexicons based on expert knowledge and the Web resources; develop a feature extraction approach that draws out character, part-of-speech, and lexical features from QA data; and improve the performance of answer entity extraction by leveraging equivalence conjunctions punctuation marks and conjunctional words in Chinese to capture dependencies between tags of entities. Experiments on question and answer entity extraction demonstrate that the Precision, Recall and F-1 score are improved using our proposed features, and the Precision and F-1 score can be further improved by considering equivalence conjunctions.

Purpose Retrieving sufficient relevant information online is difficult for many people because they use too few keywords to search and search engines do not provide many support tools. To further complicate the search, users often ignore support tools when available. Our goal is to evaluate in a realistic setting when users use support tools and how they perceive these tools. Methods We compared three medical search engines with support tools that require more or less effort from users to form a query and evaluate results. We carried out an end user study with 23 users who were asked to find information, i.e., subtopics and supporting abstracts, for a given theme. We used a balanced within-subjects design and report on the effectiveness, efficiency and usability of the support tools from the end user perspective. Conclusions We found significant differences in efficiency but did not find significant differences in effectiveness between the three search engines. Dynamic user support tools requiring less effort led to higher efficiency. Fewer searches were needed and more documents were found per search when both query reformulation and result review tools dynamically adjust to the user query. The query reformulation tool that provided a long list of keywords, dynamically adjusted to the user query, was used most often and led to more subtopics. As hypothesized, the dynamic result review tools were used more often and led to more subtopics than static ones. These results were corroborated by the usability questionnaires, which showed that support tools that dynamically optimize output were preferred.

Link (association) analysis has been used in the criminal justice domain to search large datasets for associations between crime entities in order to facilitate crime investigations. However, link analysis still faces many challenging problems, such as information overload, high search complexity, and heavy reliance on domain knowledge. To address these challenges, this article proposes several techniques for automated, effective, and efficient link analysis. These techniques include the co-occurrence analysis, the shortest path algorithm, and a heuristic approach to identifying associations and determining their importance. We developed a prototype system called CrimeLink Explorer based on the proposed techniques. Results of a user study with 10 crime investigators from the Tucson Police Department showed that our system could help subjects conduct link analysis more efficiently than traditional single-level link analysis tools. Moreover, subjects believed that association paths found based on the heuristic approach were more accurate than those found based solely on the co-occurrence analysis and that the automated link analysis system would be of great help in crime investigations.

Identity deception, specifically identity concealment, is a serious problem encountered in the law enforcement and intelligence communities. In this paper, the authors discuss techniques that can automatically detect identity deception. Most of the existing techniques are experimental and cannot be easily applied to real applications because of problems such as missing values and large data size. The authors propose an adaptive detection algorithm that adapts well to incomplete identities with missing values and to large datasets containing millions of records. The authors describe three experiments to show that the algorithm is significantly more efficient than the existing record comparison algorithm with little loss in accuracy. It can identify deception having incomplete identities with high precision. In addition, it demonstrates excellent efficiency and scalability for large databases. A case study conducted in another law enforcement agency shows that the authors' algorithm is useful in detecting both intentional deception and unintentional data errors

A new generation of data mining tools and applications work to unearth hidden patterns in large volumes of crime data.

Knowledge about the structure and organization of criminal networks is important for both crime investigation and the development of effective strategies to prevent crimes. However, except for network visualization, criminal network analysis remains primarily a manual process. Existing tools do not provide advanced structural analysis techniques that allow extraction of network knowledge from large volumes of criminal-justice data. To help law enforcement and intelligence agencies discover criminal network knowledge efficiently and effectively, in this research we proposed a framework for automated network analysis and visualization. The framework included four stages: network creation, network partition, structural analysis, and network visualization. Based upon it, we have developed a system called CrimeNet Explorer that incorporates several advanced techniques: a concept space approach, hierarchical clustering, social network analysis methods, and multidimensional scaling. Results from controlled experiments involving student subjects demonstrated that our system could achieve higher clustering recall and precision than did untrained subjects when detecting subgroups from criminal networks. Moreover, subjects identified central members and interaction patterns between groups significantly faster with the help of structural analysis functionality than with only visualization functionality. No significant gain in effectiveness was present, however. Our domain experts also reported that they believed CrimeNet Explorer could be very useful in crime investigation.

Effective and efficient link analysis techniques are needed to help law enforcement and intelligence agencies fight organized crimes such as narcotics violation, terrorism, and kidnapping. In this paper, we propose a link analysis technique that uses shortest-path algorithms, priority-first-search (PFS) and two-tree PFS, to identify the strongest association paths between entities in a criminal network. To evaluate effectiveness, we compared the PFS algorithms with crime investigators' typical association-search approach, as represented by a modified breadth-first-search (BFS). Our domain expert considered the association paths identified by PFS algorithms to be useful about 70% of the time, whereas the modified BFS algorithm's precision rates were only 30% for a kidnapping network and 16.7% for a narcotics network. Efficiency of the two-tree PFS was better for a small, dense kidnapping network, and the PFS was better for the large, sparse narcotics network.

P2P lending directly connects borrowers and lenders without a financial institution as the intermediary. This new form of crowdfunding brings lenders more investment opportunities, but also poses unprecedented risks of default and fraud. This research-in-progress paper focuses on a specific type of fraud, loan request fraud, which may be unique to lenders on Chinese P2P lending sites due to the lack of nationwide credit rating systems in China. We propose research questions surrounding the problem of loan request fraud (its types, features, and detection methods) and present our research methodology and project plans. Specifically, we plan to develop data mining based methods and employ a big data approach to address our research questions. With the help of large volumes of data from a variety of sources, we will be able to find ways to leverage rich datasets about user behaviors and transaction histories to detect loan request fraud more effectively and efficiently.

Experienced users learn the “tricks of the trade” that allow them to make effective and efficient use of enterprise systems. New users and even experienced ones accessing unfamiliar parts of the system, however, are stymied by the lack of navigational cues, the ubiquity of tabbed pages, and the sheer number of fields that may be required. In earlier work, we documented the usability issues users are facing in the field, derived four design principles for guiding the design of systems that collaborate with their users, and implemented a database framework for supporting those principles. In this paper, we present two interventions targeted at the design principle that calls for providing navigational and progress guidance, while supporting the principle requiring that the system make use of contextual information in presenting data and choices to the user. Using visualizations in conjunction with a playback mechanism, users can learn about system-supported processes, find detailed information on the particular process instances on which they are working, and view automated playbacks of users completing the same and other types of tasks. This paper is the next step in designing enterprise systems that help, rather than hinder, their human partners by sharing their knowledge.

We present a proof-of-concept prototype that demonstrates novel collaborative features for supporting enterprise system users. The purpose of the prototype is to illustrate the implementation of design principles derived from field studies of Enterprise Resource Planning system (ERP) users. Its novel features are designed to improve the users’ understanding of the business context of their interactions with the system and help them learn to operate it by viewing how tasks have been performed in the past.

Enterprise Resource Planning (ERP) systems have received well-deserved attention from both the industry and the research communities, yet the usability of these systems remains a challenge. This paper proposes a model of the usability of ERP systems based on collaboration theory. In this model, Commitment to Joint Activity (CJA), Mutual Responsiveness (MR), and Commitment to Mutual Support (CMS) are properties of Shared Cooperative Activity between the user and the ERP system. The strength of this collaboration (conceptualized as the interaction of the individual properties: CJA, MR, and CMS) impacts the system's usability. The proposed constructs are illustrated with data from a field-based case study. Our findings provide initial support for our model of human-computer collaboration in ERP systems.

Decision-making in many application domains needs to take into consideration of some sorts of networks. Examples include e-commerce and marketing [6, 10], strategic planning [21], knowledge management [12], and Web mining [5, 13]. Since the late 1990s a large number of articles have been published in Nature, Science, and other leading journals in many disciplines, proposing new network models, techniques, and applications (e.g., [3, 22, 25]). This trend has been accompanied by the increasing popularity of social networking sites such as FaceBook and MySpace. As a result, research on social network data mining, or simply network mining, has attracted much attention from both academics and practitioners.

This research is motivated by data mining for wireless sensor network applications. The authors consider applications where data is acquired in real-time, and thus data mining is performed on live streams of data rather than on stored databases. One challenge in supporting such applications is that sensor node power is a precious resource that needs to be managed as such. To conserve energy in the sensor field, the authors propose and evaluate several approaches to acquiring, and then caching data in a sensor field data server. The authors show that for true real-time applications, for which response time dictates data quality, policies that emulate cache hits by computing and returning approximate values for sensor data yield a simultaneous quality improvement and cost saving. This “win-win” is because when data acquisition response time is sufficiently important, the decrease in resource consumption and increase in data quality achieved by using approximate values outweighs the negative impact on data accuracy due to the approximation. In contrast, when data accuracy drives quality, a linear trade-off between resource consumption and data accuracy emerges. The authors then identify caching and lookup policies for which the sensor field query rate is bounded when servicing an arbitrary workload of user queries. This upper bound is achieved by having multiple user queries share the cost of a sensor field query. Finally, the authors discuss the challenges facing sensor network data mining applications in terms of data collection, warehousing, and mining techniques.

In recent years terrorist groups have been using the World-Wide Web to spread their ideologies, disseminate propaganda, and recruit members. Studying the terrorist Web sites may help us understand the characteristics of these Web sites and predict terrorist activities. In this chapter, we propose to apply network topological analysis methods on systematically collected the terrorist Web site data and to study the structural characteristics at the Web page level. We conducted a case study using the methods on three collections of terrorist Web sites: Middle- Eastern, US domestic, and Latin-American. We found that the Web page networks from these three collections have the small-world and scale-free characteristics. We also found that smaller size Web sites which share similar interests tend to make stronger inter-site linkages, which help them form the giant component in the networks.

Automatically extracting gene-pathway relations from medical research texts gives researchers access to the latest findings in a structured format. Such relations must be precise to be useful. We present two case studies of approaches used to automatically extract gene-pathway relations from text. Each technique has performed at or near the 90 percent precision level making them good candidates to perform the extraction task. In addition, we present a visualization system that uses XML to interface with the extracted gene-pathway relations. The user-selected relations are automatically presented in a network display, inspired by the pathway maps created by gene researchers manually. Future research involves identification of equivalent relations expressed differently by authors and identification of relations that contradict each other along with the inquiry of how this information is useful to researchers.

In recent years, terrorist groups have used the WWW to spread their ideologies, disseminate propaganda, and recruit members. Studying the terrorist websites may help us understand the characteristics of these websites and predict terrorist activities. In this paper, we propose to apply network topological analysis methods on systematically collected the terrorist website data and to study the structural characteristics at the Web page level. We conducted a case study using the methods on three collections of Middle-Eastern, US domestic, and Latin-American terrorist websites. We found that these three networks have the small-world and scale-free characteristics. We also found that smaller size websites which share same interests tend to make stronger inter-website linkage relationships.

The goal of the COPLINK project is to develop information and knowledge management systems technologies and methodology appropriate for capturing, accessing, analyzing, visualizing, and sharing law enforcement related information. Continuing our pursuit of this goal, we have developed a framework for creating cross-jurisdictional Criminal Activity Networks for social network analysis, and have constructed a taxonomy of criminal identity resolution problems for detecting intentional identity deception and unintentional identity errors.

It is very important for us to understand the functions and structures of terrorist networks to win the battle against terror. However, previous studies of terrorist network structure have generated little actionable results. This is mainly due to the difficulty in collecting and accessing reliable data and the lack of advanced network analysis methodologies in the field. To address these problems, we employed several advance network analysis techniques ranging from social network analysis to Web structural mining on a Global Salafi Jihad network dataset collected through a large scale empirical study. Our study demonstrated the effectiveness and usefulness of advanced network techniques in terrorist network analysis domain. We also introduced the Web structural mining technique into the terrorist network analysis field which, to the best our knowledge, has never been used in this domain. More importantly, the results from our analysis provide not only insights for terrorism research community but also empirical implications that may help law-reinforcement, intelligence, and security communities to make our nation safer.

The BorderSafe project funded by Department of Homeland Security (DHS) and the Corporation for National Research Initiatives (CNRI) aims to develop, foster, and leverage information sharing between law enforcement agencies for border safety and national security. The partners in the project include the Artificial Intelligence (AI) Lab at the University of Arizona, Tucson Police Department (TPD), Pima County Sheriff's Department (PCSD), Tucson Customs and Border Protection (CBP), San Diego Automated Regional Justice Information System (ARJIS), and the San Diego Supercomputer Center (SDSC). We describe the three major areas of research in the BorderSafe project at the AI Lab, University of Arizona.

Ever since the 9–11 incident, the multidisciplinary field of terrorism has experienced tremendous growth. As the domain has benefited greatly from recent advances in information technologies, more complex and challenging new issues have emerged from numerous counter-terrorism-related research communities as well as governments of all levels. In this paper, we describe an advanced knowledge discovery approach to addressing terrorism threats. We experimented with our approach in a project called Terrorism Knowledge Discovery Project that consists of several custom-built knowledge portals. The main focus of this project is to provide advanced methodologies for analyzing terrorism research, terrorists, and the terrorized groups (victims). Once completed, the system can also become a major learning resource and tool that the general community can use to heighten their awareness and understanding of global terrorism phenomenon, to learn how best they can respond to terrorism and, eventually, to garner significant grass root support for the government’s efforts to keep America safe.

Dynamic criminal network analysis is important for national security but also very challenging. However, little research has been done in this area. In this paper we propose to use several descriptive measures from social network analysis research to help detect and describe changes in criminal organizations. These measures include centrality for individuals, and density, cohesion, and stability for groups. We also employ visualization and animation methods to present the evolution process of criminal networks. We conducted a field study with several domain experts to validate our findings from the analysis of the dynamics of a narcotics network. The feedback from our domain experts showed that our approaches and the prototype system could be very helpful for capturing the dynamics of criminal organizations and assisting crime investigation and criminal prosecution.

Expertise management systems are being widely adopted in organizations to manage tacit knowledge embedded in employees' heads. These systems have successfully applied many information technologies developed in fields such as information retrieval and document management to support expertise information collection, processing, and distribution. In this paper, we investigate the potentials of applying visualization techniques to support exploration of an expertise space. We implemented two widely applied dimensionality reduction visualization techniques, the self-organizing map and multidimensional scaling; to generate expert map and expertise field map visualizations based on an expertise data set. Our proposed approach is generic for automatic mapping of expertise space of an organization, research field, scientific domain, etc. Our initial analysis on the visualization results indicated that the expert map and expertise field map captured useful underlying structures of the expertise space and had the potential to support more efficient and effective expertise information searching and browsing.

Knowledge about criminal networks has important implications for crime investigation and the anti-terrorism campaign. However, lack of advanced, automated techniques has limited law enforcement and intelligence agencies' ability to combat crime by discovering structural patterns in criminal networks. In this research we used the concept space approach, clustering technology, social network analysis measures and approaches, and multidimensional scaling methods for automatic extraction, analysis, and visualization of criminal networks and their structural patterns. We conducted a case study with crime investigators from the Tucson Police Department. They validated the structural patterns discovered from gang and narcotics criminal enterprises. The results showed that the approaches we proposed could detect subgroups, central members, and between-group interaction patterns correctly most of the time. Moreover, our system could extract the overall structure for a network that might be useful in the development of effective disruptive strategies for criminal networks.

Link (association) analysis has been used in law enforcement and intelligence domains to extract and search associations between people from large datasets. Nonetheless, link analysis still faces many challenging problems, such as information overload, high search complexity, and heavy reliance on domain knowledge. To address these challenges and enable crime investigators to conduct automated, effective, and efficient link analysis, we proposed three techniques which include: the concept space approach, a shortest-path algorithm, and a heuristic approach that captures domain knowledge for determining importance of associations. We implemented a system called CrimeLink Explorer based on the proposed techniques. Results from our user study involving ten crime investigators from the Tucson Police Department showed that our system could help subjects conduct link analysis more efficiently. Additionally, subjects concluded that association paths found based on the heuristic approach were more accurate than those found based on the concept space approach.