ACM Transactions on Knowledge Discovery from Data

Published by Association for Computing Machinery
Print ISSN: 1556-4681
Spatial and temporal view of Drosophila images across different stages (1–16) of development (of the same gene Kr). The textural features (based on the morphology of the embryo) are different from the gene expression, which is indicted by the blue staining.  
Plot of discriminant features corresponding to stage range 4–6 (left graph) and stage range 7– 8 (right graph), where rectangular blocks indicate the regions on the images with nonzero coefficients.  
Gene expression in a developing embryo occurs in particular cells (spatial patterns) in a time-specific manner (temporal patterns), which leads to the differentiation of cell fates. Images of a Drosophila melanogaster embryo at a given developmental stage, showing a particular gene expression pattern revealed by a gene-specific probe, can be compared for spatial overlaps. The comparison is fundamentally important to formulating and testing gene interaction hypotheses. Expression pattern comparison is most biologically meaningful when images from a similar time point (developmental stage) are compared. In this paper, we present LdaPath, a novel formulation of Linear Discriminant Analysis (LDA) for automatic developmental stage range classification. It employs multivariate linear regression with the L(1)-norm penalty controlled by a regularization parameter for feature extraction and visualization. LdaPath computes an entire solution path for all values of regularization parameter with essentially the same computational cost as fitting one LDA model. Thus, it facilitates efficient model selection. It is based on the equivalence relationship between LDA and the least squares method for multi-class classifications. This equivalence relationship is established under a mild condition, which we show empirically to hold for many high-dimensional datasets, such as expression pattern images. Our experiments on a collection of 2705 expression pattern images show the effectiveness of the proposed algorithm. Results also show that the LDA model resulting from LdaPath is sparse, and irrelevant features may be removed. Thus, LdaPath provides a general framework for simultaneous feature selection and feature extraction.
BACKGROUND: We recently described "Author-ity," a model for estimating the probability that two articles in MEDLINE, sharing the same author name, were written by the same individual. Features include shared title words, journal name, coauthors, medical subject headings, language, affiliations, and author name features (middle initial, suffix, and prevalence in MEDLINE). Here we test the hypothesis that the Author-ity model will suffice to disambiguate author names for the vast majority of articles in MEDLINE. METHODS: Enhancements include: (a) incorporating first names and their variants, email addresses, and correlations between specific last names and affiliation words; (b) new methods of generating large unbiased training sets; (c) new methods for estimating the prior probability; (d) a weighted least squares algorithm for correcting transitivity violations; and (e) a maximum likelihood based agglomerative algorithm for computing clusters of articles that represent inferred author-individuals. RESULTS: Pairwise comparisons were computed for all author names on all 15.3 million articles in MEDLINE (2006 baseline), that share last name and first initial, to create Author-ity 2006, a database that has each name on each article assigned to one of 6.7 million inferred author-individual clusters. Recall is estimated at ~98.8%. Lumping (putting two different individuals into the same cluster) affects ~0.5% of clusters, whereas splitting (assigning articles written by the same individual to >1 cluster) affects ~2% of articles. IMPACT: The Author-ity model can be applied generally to other bibliographic databases. Author name disambiguation allows information retrieval and data integration to become person-centered, not just document-centered, setting the stage for new data mining and social network tools that will facilitate the analysis of scholarly publishing and collaboration behavior. AVAILABILITY: The Author-ity 2006 database is available for nonprofit academic research, and can be freely queried via
In this article, we propose a methodology for identifying predictive physiological patterns in the absence of prior knowledge. We use the principle of conservation to identify activity that consistently precedes an outcome in patients, and describe a two-stage process that allows us to efficiently search for such patterns in large datasets. This involves first transforming continuous physiological signals from patients into symbolic sequences, and then searching for patterns in these reduced representations that are strongly associated with an outcome. Our strategy of identifying conserved activity that is unlikely to have occurred purely by chance in symbolic data is analogous to the discovery of regulatory motifs in genomic datasets. We build upon existing work in this area, generalizing the notion of a regulatory motif and enhancing current techniques to operate robustly on non-genomic data. We also address two significant considerations associated with motif discovery in general: computational efficiency and robustness in the presence of degeneracy and noise. To deal with these issues, we introduce the concept of active regions and new subset-based techniques such as a two-layer Gibbs sampling algorithm. These extensions allow for a framework for information inference, where precursors are identified as approximately conserved activity of arbitrary complexity preceding multiple occurrences of an event. We evaluated our solution on a population of patients who experienced sudden cardiac death and attempted to discover electrocardiographic activity that may be associated with the endpoint of death. To assess the predictive patterns discovered, we compared likelihood scores for motifs in the sudden death population against control populations of normal individuals and those with non-fatal supraventricular arrhythmias. Our results suggest that predictive motif discovery may be able to identify clinically relevant information even in the absence of significant prior knowledge.
A demonstration for our motivation and learning process. The shared group structure matrix is jointly determined by the rating graph and trust graph. The rating matrix contains two groups of users' reviews about movies, where a smile face represents satisfactory and an angry face represents unsatisfactory. The trust matrix contains users' trust evaluation toward other users, where 1 represents trust and 0 represents distrust. The question mark represents missing value in both graphs. The 1s in the cluster information matrix indicate users are in the corresponding group, while 0s represent users are not in that group. 
Along with increasing popularity of social websites, online users rely more on the trustworthiness information to make decisions, extract and filter information, and tag and build connections with other users. However, such social network data often suffer from severe data sparsity and are not able to provide users with enough information. Therefore, trust prediction has emerged as an important topic in social network research. Traditional approaches are primarily based on exploring trust graph topology itself. However, research in sociology and our life experience suggest that people who are in the same social circle often exhibit similar behaviors and tastes. To take advantage of the ancillary information for trust prediction, the challenge then becomes what to transfer and how to transfer. In this article, we address this problem by aggregating heterogeneous social networks and propose a novel joint social networks mining (JSNM) method. Our new joint learning model explores the user-group-level similarity between correlated graphs and simultaneously learns the individual graph structure; therefore, the shared structures and patterns from multiple social networks can be utilized to enhance the prediction tasks. As a result, we not only improve the trust prediction in the target graph but also facilitate other information retrieval tasks in the auxiliary graphs. To optimize the proposed objective function, we use the alternative technique to break down the objective function into several manageable subproblems. We further introduce the auxiliary function to solve the optimization problems with rigorously proved convergence. The extensive experiments have been conducted on both synthetic and real- world data. All empirical results demonstrate the effectiveness of our method.
Statistics of the benchmark data sets. 
Illustration of the transformation matrix Z in Eq. (1), where P denotes the sparse component with the zero-value entries represented by white blocks, and Q denotes the low-rank component. 
Performance comparison of six multi-task learning algorithms with different training ratios in terms of average AUC (left plot), Macro F1 (middle plot), and Micro F1 (right plot). The index on x-axis corresponds to the training ratio varying from 0.1 to 0.9. 
We consider the problem of learning incoherent sparse and low-rank patterns from multiple tasks. Our approach is based on a linear multi-task learning formulation, in which the sparse and low-rank patterns are induced by a cardinality regularization term and a low-rank constraint, respectively. This formulation is non-convex; we convert it into its convex surrogate, which can be routinely solved via semidefinite programming for small-size problems. We propose to employ the general projected gradient scheme to efficiently solve such a convex surrogate; however, in the optimization formulation, the objective function is non-differentiable and the feasible domain is non-trivial. We present the procedures for computing the projected gradient and ensuring the global convergence of the projected gradient scheme. The computation of projected gradient involves a constrained optimization problem; we show that the optimal solution to such a problem can be obtained via solving an unconstrained optimization subproblem and an Euclidean projection subproblem. We also present two projected gradient algorithms and analyze their rates of convergence in details. In addition, we illustrate the use of the presented projected gradient algorithms for the proposed multi-task learning formulation using the least squares loss. Experimental results on a collection of real-world data sets demonstrate the effectiveness of the proposed multi-task learning formulation and the efficiency of the proposed projected gradient algorithms.
Traditional anomaly detection on social media mostly focuses on individual point anomalies while anomalous phenomena usually occur in groups. Therefore it is valuable to study the collective behavior of individuals and detect group anomalies. Existing group anomaly detection approaches rely on the assumption that the groups are known, which can hardly be true in real world social media applications. In this paper, we take a generative approach by proposing a hierarchical Bayes model: Group Latent Anomaly Detection (GLAD) model. GLAD takes both pair-wise and point-wise data as input, automatically infers the groups and detects group anomalies simultaneously. To account for the dynamic properties of the social media data, we further generalize GLAD to its dynamic extension d-GLAD. We conduct extensive experiments to evaluate our models on both synthetic and real world datasets. The empirical results demonstrate that our approach is effective and robust in discovering latent groups and detecting group anomalies.
The setup for the network alignment problem. The goal is to maximize the number of squares in any matching while maximizing the weight of the matching as well.
A small sample problem and the data for the QP formulation.
The graph (b) is the factor-graph representation of the problem in (a).
Network alignment generalizes and unifies several approaches for forming a matching or alignment between the vertices of two graphs. We study a mathematical programming framework for network alignment problem and a sparse variation of it where only a small number of matches between the vertices of the two graphs are possible. We propose a new message passing algorithm that allows us to compute, very efficiently, approximate solutions to the sparse network alignment problems with graph sizes as large as hundreds of thousands of vertices. We also provide extensive simulations comparing our algorithms with two of the best solvers for network alignment problems on two synthetic matching problems, two bioinformatics problems, and three large ontology alignment problems including a multilingual problem with a known labeled alignment.
Collaborative tagging systems, such as Delicious, CiteULike, and others, allow users to annotate resources, e.g., Web pages or scientific papers, with descriptive labels called tags. The social annotations contributed by thousands of users, can potentially be used to infer categorical knowledge, classify documents or recommend new relevant information. Traditional text inference methods do not make best use of social annotation, since they do not take into account variations in individual users' perspectives and vocabulary. In a previous work, we introduced a simple probabilistic model that takes interests of individual annotators into account in order to find hidden topics of annotated resources. Unfortunately, that approach had one major shortcoming: the number of topics and interests must be specified a priori. To address this drawback, we extend the model to a fully Bayesian framework, which offers a way to automatically estimate these numbers. In particular, the model allows the number of interests and topics to change as suggested by the structure of the data. We evaluate the proposed model in detail on the synthetic and real-world data by comparing its performance to Latent Dirichlet Allocation on the topic extraction task. For the latter evaluation, we apply the model to infer topics of Web resources from social annotations obtained from Delicious in order to discover new resources similar to a specified one. Our empirical results demonstrate that the proposed model is a promising method for exploiting social knowledge contained in user-generated annotations. Comment: 29 Pages, Accepted for publication at ACM Transactions on Knowledge Discovery from Data(TKDD) on March 2, 2010
In many applications, an anomaly detection system presents the most anomalous data instance to a human analyst, who then must determine whether the instance is truly of interest (e.g. a threat in a security setting). Unfortunately, most anomaly detectors provide no explanation about why an instance was considered anomalous, leaving the analyst with no guidance about where to begin the investigation. To address this issue, we study the problems of computing and evaluating sequential feature explanations (SFEs) for anomaly detectors. An SFE of an anomaly is a sequence of features, which are presented to the analyst one at a time (in order) until the information contained in the highlighted features is enough for the analyst to make a confident judgement about the anomaly. Since analyst effort is related to the amount of information that they consider in an investigation, an explanation's quality is related to the number of features that must be revealed to attain confidence. One of our main contributions is to present a novel framework for large scale quantitative evaluations of SFEs, where the quality measure is based on analyst effort. To do this we construct anomaly detection benchmarks from real data sets along with artificial experts that can be simulated for evaluation. Our second contribution is to evaluate several novel explanation approaches within the framework and on traditional anomaly detection benchmarks, offering several insights into the approaches.
Ensemble learning for anomaly detection has been barely studied, due to difficulty in acquiring ground truth and the lack of inherent objective functions. In contrast, ensemble approaches for classification and clustering have been studied and effectively used for long. Our work taps into this gap and builds a new ensemble approach for anomaly detection, with application to event detection in temporal graphs as well as outlier detection in no-graph settings. It handles and combines multiple heterogeneous detectors to yield improved and robust performance. Importantly, trusting results from all the constituent detectors may deteriorate the overall performance of the ensemble, as some detectors could provide inaccurate results depending on the type of data in hand and the underlying assumptions of a detector. This suggests that combining the detectors selectively is key to building effective anomaly ensembles—hence “less is more”. In this paper we propose a novel ensemble approach called SELECT for anomaly detection, which automatically and systematically selects the results from constituent detectors to combine in a fully unsupervised fashion. We apply our method to event detection in temporal graphs and outlier detection in multi-dimensional point data (no-graph), where SELECT successfully utilizes five base detectors and seven consensus methods under a unified ensemble framework. We provide extensive quantitative evaluation of our approach for event detection on five real-world datasets (four with ground truth events), including Enron email communications, RealityMining SMS and phone call records, New York Times news corpus, and World Cup 2014 Twitter news feed. We also provide results for outlier detection on seven real-world multi-dimensional point datasets from UCI Machine Learning Repository. Thanks to its selection mechanism, SELECT yields superior performance compared to the individual detectors alone, the full ensemble (naively combining all results), an existing diversity-based ensemble, and an existing weighted ensemble approach.
Global distribution 
Group based anonymization is the most widely studied approach for privacy preserving data publishing. This includes k-anonymity, l-diversity, and t-closeness, to name a few. The goal of this paper is to raise a fundamental issue on the privacy exposure of the current group based approach. This has been overlooked in the past. The group based anonymization approach basically hides each individual record behind a group to preserve data privacy. If not properly anonymized, patterns can actually be derived from the published data and be used by the adversary to breach individual privacy. For example, from the medical records released, if patterns such as people from certain countries rarely suffer from some disease can be derived, then the information can be used to imply linkage of other people in an anonymized group with this disease with higher likelihood. We call the derived patterns from the published data the foreground knowledge. This is in contrast to the background knowledge that the adversary may obtain from other channels as studied in some previous work. Finally, we show by experiments that the attack is realistic in the privacy benchmark dataset under the traditional group based anonymization approach. Comment: 11 pages
Some existing notions of redundancy among association rules allow for a logical-style characterization and lead to irredundant bases of absolutely minimum size. We push the intuition of redundancy further to find an intuitive notion of novelty of an association rule, with respect to other rules. Namely, an irredundant rule is so because its confidence is higher than what the rest of the rules would suggest; then, one can ask: how much higher? We propose to measure such a sort of novelty through the confidence boost of a rule. Acting as a complement to confidence and support, the confidence boost helps to obtain small and crisp sets of mined association rules and solves the well-known problem that, in certain cases, rules of negative correlation may pass the confidence bound. We analyze the properties of two versions of the notion of confidence boost, one of them a natural generalization of the other. We develop algorithms to filter rules according to their confidence boost, compare the concept to some similar notions in the literature, and describe the results of some experimentation employing the new notions on standard benchmark datasets. We describe an open source association mining tool that embodies one of our variants of confidence boost in such a way that the data mining process does not require the user to select any value for any parameter.
The tasks of extracting (top-$K$) Frequent Itemsets (FI's) and Association Rules (AR's) are fundamental primitives in data mining and database applications. Exact algorithms for these problems exist and are widely used, but their running time is hindered by the need of scanning the entire dataset, possibly multiple times. High quality approximations of FI's and AR's are sufficient for most practical uses, and a number of recent works explored the application of sampling for fast discovery of approximate solutions to the problems. However, these works do not provide satisfactory performance guarantees on the quality of the approximation, due to the difficulty of bounding the probability of under- or over-sampling any one of an unknown number of frequent itemsets. In this work we circumvent this issue by applying the statistical concept of \emph{Vapnik-Chervonenkis (VC) dimension} to develop a novel technique for providing tight bounds on the sample size that guarantees approximation within user-specified parameters. Our technique applies both to absolute and to relative approximations of (top-$K$) FI's and AR's. The resulting sample size is linearly dependent on the VC-dimension of a range space associated with the dataset to be mined. The main theoretical contribution of this work is a proof that the VC-dimension of this range space is upper bounded by an easy-to-compute characteristic quantity of the dataset which we call \emph{d-index}, and is the maximum integer $d$ such that the dataset contains at least $d$ transactions of length at least $d$ such that no one of them is a superset of or equal to another. We show that this bound is strict for a large class of datasets.
Real-world networks are often organized as modules or communities of similar nodes that serve as functional units. These networks are also rich in content, with nodes having distinguishing features or attributes. In order to discover a network's modular structure, it is necessary to take into account not only its links but also node attributes. We describe an information-theoretic method that identifies modules by compressing descriptions of information flow on a network. Our formulation introduces node content into the description of information flow, which we then minimize to discover groups of nodes with similar attributes that also tend to trap the flow of information. The method has several advantages: it is conceptually simple and does not require ad-hoc parameters to specify the number of modules or to control the relative contribution of links and node attributes to network structure. We apply the proposed method to partition real-world networks with known community structure. We demonstrate that adding node attributes helps recover the underlying community structure in content-rich networks more effectively than using links alone. In addition, we show that our method is faster and more accurate than alternative state-of-the-art algorithms.
People's personal social networks are big and cluttered, and currently there is no good way to automatically organize them. Social networking sites allow users to manually categorize their friends into social circles (e.g. 'circles' on Google+, and 'lists' on Facebook and Twitter), however they are laborious to construct and must be updated whenever a user's network grows. In this paper, we study the novel task of automatically identifying users' social circles. We pose this task as a multi-membership node clustering problem on a user's ego-network, a network of connections between her friends. We develop a model for detecting circles that combines network structure as well as user profile information. For each circle we learn its members and the circle-specific user profile similarity metric. Modeling node membership to multiple circles allows us to detect overlapping as well as hierarchically nested circles. Experiments show that our model accurately identifies circles on a diverse set of data from Facebook, Google+, and Twitter, for all of which we obtain hand-labeled ground-truth.
With the advancement of information systems, means of communications are becoming cheaper, faster and more available. Today, millions of people carrying smart-phones or tablets are able to communicate at practically any time and anywhere they want. Among others, they can access their e-mails, comment on weblogs, watch and post comments on videos, make phone calls or text messages almost ubiquitously. Given this scenario, in this paper we tackle a fundamental aspect of this new era of communication: how the time intervals between communication events behave for different technologies and means of communications? Are there universal patterns for the inter-event time distribution (IED)? In which ways inter-event times behave differently among particular technologies? To answer these questions, we analyze eight different datasets from real and modern communication data and we found four well defined patterns that are seen in all the eight datasets. Moreover, we propose the use of the Self-Feeding Process (SFP) to generate inter-event times between communications. The SFP is extremely parsimonious point process that requires at most two parameters and is able to generate inter-event times with all the universal properties we observed in the data. We show the potential application of SFP by proposing a framework to generate a synthetic dataset containing realistic communication events of any one of the analyzed means of communications (e.g. phone calls, e-mails, comments on blogs) and an algorithm to detect anomalies.
Information diffusion and virus propagation are fundamental processes taking place in networks. While it is often possible to directly observe when nodes become infected with a virus or publish the information, observing individual transmissions (who infects whom, or who influences whom) is typically very difficult. Furthermore, in many applications, the underlying network over which the diffusions and propagations spread is actually unobserved. We tackle these challenges by developing a method for tracing paths of diffusion and influence through networks and inferring the networks over which contagions propagate. Given the times when nodes adopt pieces of information or become infected, we identify the optimal network that best explains the observed infection times. Since the optimization problem is NP-hard to solve exactly, we develop an efficient approximation algorithm that scales to large datasets and finds provably near-optimal networks. We demonstrate the effectiveness of our approach by tracing information diffusion in a set of 170 million blogs and news articles over a one year period to infer how information flows through the online media space. We find that the diffusion network of news for the top 1,000 media sites and blogs tends to have a core-periphery structure with a small set of core media sites that diffuse information to the rest of the Web. These sites tend to have stable circles of influence with more general news media sites acting as connectors between them.
Feature Selection is an important technique in machine learning and pattern classification, especially for handling high-dimensional data. Most existing studies have been restricted to batch learning, which is often inefficient and poorly scalable when handling big data in real world, especially when data arrives sequentially. Recent years have witnessed some emerging feature selection techniques using online learning. Despite enjoying significant advantages in efficiency and scalability, the existing online feature selection methods are not always accurate enough, and still not sufficiently fast when handling massive-scale data with ultra-high dimensionality. To address the limitations, we propose a novel online feature selection method by exploiting second-order information with optimized implementations, which not only improves the learning efficacy, but also significantly enhances computational efficiency. We conduct extensive experiments for evaluating both learning accuracy and time cost of different algorithms on massive-scale synthetic and real-world datasets, including a dataset with billion-scale features. Our results show that our technique achieves highly competitive accuracy as compared with state-of-the-art batch feature selection methods, but consumes significantly low computational cost that is orders of magnitude lower than both state-of-the-art batch and online feature selection methods. On a billion-scale synthetic dataset (1-billion dimensions, 1-billion nonzero features, and 1-million samples), our algorithm took only eight minutes with a normal single machine.
The data in many disciplines such as social networks, web analysis, etc. is link-based, and the link structure can be exploited for many different data mining tasks. In this paper, we consider the problem of temporal link prediction: Given link data for times 1 through T, can we predict the links at time T+1? If our data has underlying periodic structure, can we predict out even further in time, i.e., links at time T+2, T+3, etc.? In this paper, we consider bipartite graphs that evolve over time and consider matrix- and tensor-based methods for predicting future links. We present a weight-based method for collapsing multi-year data into a single matrix. We show how the well-known Katz method for link prediction can be extended to bipartite graphs and, moreover, approximated in a scalable way using a truncated singular value decomposition. Using a CANDECOMP/PARAFAC tensor decomposition of the data, we illustrate the usefulness of exploiting the natural three-dimensional structure of temporal link data. Through several numerical experiments, we demonstrate that both matrix- and tensor-based techniques are effective for temporal link prediction despite the inherent difficulty of the problem. Additionally, we show that tensor-based techniques are particularly effective for temporal data with varying periodic patterns.
Spectrum of Computational Models for network sampling: from static to streaming. 
Network sampling is integral to the analysis of social, information, and biological networks. Since many real-world networks are massive in size, continuously evolving, and/or distributed in nature, the network structure is often sampled in order to facilitate study. For these reasons, a more thorough and complete understanding of network sampling is critical to support the field of network science. In this paper, we outline a framework for the general problem of network sampling, by highlighting the different objectives, population and units of interest, and classes of network sampling methods. In addition, we propose a spectrum of computational models for network sampling methods, ranging from the traditionally studied model based on the assumption of a static domain to a more challenging model that is appropriate for streaming domains. We design a family of sampling methods based on the concept of graph induction that generalize across the full spectrum of computational models (from static to streaming) while efficiently preserving many of the topological properties of the input graphs. Furthermore, we demonstrate how traditional static sampling algorithms can be modified for graph streams for each of the three main classes of sampling methods: node, edge, and topology-based sampling. Our experimental results indicate that our proposed family of sampling methods more accurately preserves the underlying properties of the graph for both static and streaming graphs. Finally, we study the impact of network sampling algorithms on the parameter estimation and performance evaluation of relational classification algorithms.
ǫ out as a function of r in the Hapmap-HGDP dataset for three different random projection methods and two different classification tasks. Vertical bars indicate the standard-deviation over the ten ten-fold cross-validation experiments and the ten choices of the random projection matrices for each of the three methods.  
Synthetic data: γ increases as a function of r in all three families of matrices. See the caption of Table 1 for an explanation of µ and σ.
Results on the Techtc300 dataset, averaged over 295 data matrices using three different random projection methods. The table shows how ǫ out , γ, t rp (in seconds), and t run (in seconds) depend on r. µ and σ indicate the mean and the standard deviation of each quantity over 295 matrices, ten ten-fold cross-validation experiments, and ten choices of random projection matrices for the three methods that we investigated. 
Let X be a data matrix of rank \rho, whose rows represent n points in d-dimensional space. The linear support vector machine constructs a hyperplane separator that maximizes the 1-norm soft margin. We develop a new oblivious dimension reduction technique which is precomputed and can be applied to any input matrix X. We prove that, with high probability, the margin and minimum enclosing ball in the feature space are preserved to within \epsilon-relative error, ensuring comparable generalization as in the original space in the case of classification. For regression, we show that the margin is preserved to \epsilon-relative error with high probability. We present extensive experiments with real and synthetic data to support our theory.
We address the problem of \emph{quantification}, a supervised learning task whose goal is, given a class, to estimate the relative frequency (or \emph{prevalence}) of the class in a dataset of unlabelled items. Quantification has several applications in data and text mining, such as estimating the prevalence of positive reviews in a set of reviews of a given product, or estimating the prevalence of a given support issue in a dataset of transcripts of phone calls to tech support. So far, quantification has been addressed by learning a general-purpose classifier, counting the unlabelled items which have been assigned the class, and tuning the obtained counts according to some heuristics. In this paper we depart from the tradition of using general-purpose classifiers, and use instead a supervised learning model for \emph{structured prediction}, capable of generating classifiers directly optimized for the (multivariate and non-linear) function used for evaluating quantification accuracy. The experiments that we have run on 5500 binary high-dimensional datasets (averaging more than 14,000 documents each) show that this method is more accurate, more stable, and more efficient than existing, state-of-the-art quantification methods.
Sybil accounts are fake identities created to unfairly increase the power or resources of a single malicious user. Researchers have long known about the existence of Sybil accounts in online communities such as file-sharing systems, but have not been able to perform large scale measurements to detect them or measure their activities. In this paper, we describe our efforts to detect, characterize and understand Sybil account activity in the Renren online social network (OSN). We use ground truth provided by Renren Inc. to build measurement based Sybil account detectors, and deploy them on Renren to detect over 100,000 Sybil accounts. We study these Sybil accounts, as well as an additional 560,000 Sybil accounts caught by Renren, and analyze their link creation behavior. Most interestingly, we find that contrary to prior conjecture, Sybil accounts in OSNs do not form tight-knit communities. Instead, they integrate into the social graph just like normal users. Using link creation timestamps, we verify that the large majority of links between Sybil accounts are created accidentally, unbeknownst to the attacker. Overall, only a very small portion of Sybil accounts are connected to other Sybils with social links. Our study shows that existing Sybil defenses are unlikely to succeed in today's OSNs, and we must design new techniques to effectively detect and defend against Sybil attacks.
We present an extension to Jaynes' maximum entropy principle that handles latent variables. The principle of latent maximum entropy we propose is di#erent from both Jaynes' maximum entropy principle and maximum likelihood estimation, but often yields better estimates in the presence of hidden variables and limited training data. We first show that solving for a latent maximum entropy model poses a hard nonlinear constrained optimization problem in general. However, we then show that feasible solutions to this problem can be obtained e#ciently for the special case of log-linear models---which forms the basis for an e#cient approximation to the latent maximum entropy principle. We derive an algorithm that combines expectation-maximization with iterative scaling to produce feasible log-linear solutions. This algorithm can be interpreted as an alternating minimization algorithm in the information divergence, and reveals an intimate connection between the latent maximum entropy and maximum likelihood principles.
How do we find a natural clustering of a real-world point set which contains an unknown number of clusters with different shapes, and which may be contaminated by noise? As most clustering algorithms were designed with certain assumptions ...
Interaction graphs are ubiquitous in many fields such as bioinformatics, sociology and physical sciences. There have been many studies in the literature targeted at studying and mining these graphs. However, almost all of them have studied these graphs ...
Behavioral targeting (BT) leverages historical user behavior to select the ads most relevant to users to display. The state-of-the-art of BT derives a linear Poisson regression model from fine-grained user behavioral data and predicts click-through rate ...
In this article, we study the problem of Web user profiling, which is aimed at finding, extracting, and fusing the “semantic”-based user profile from the Web. Previously, Web user profiling was often undertaken by creating a list of keywords ...
A topic taxonomy is an effective representation that describes salient features of virtual groups or online communities. A topic taxonomy consists of topic nodes. Each internal node is defined by its vertical path (i.e., ancestor and child nodes) and its horizonal list of attributes (or terms). In a text-dominant environment, a topic taxonomy can be used to flexibly describe a group's interests with varying granularity. However, the stagnant nature of a taxonomy may fail to timely capture the dynamic change of a group's interest. This article addresses the problem of how to adapt a topic taxonomy to the accumulated data that reflects the change of a group's interest to achieve dynamic group profiling. We first discuss the issues related to topic taxonomy. We next formulate taxonomy adaptation as an optimization problem to find the taxonomy that best fits the data. We then present a viable algorithm that can efficiently accomplish taxonomy adaptation. We conduct extensive experiments to evaluate our approach's efficacy for group profiling, compare the approach with some alternatives, and study its performance for dynamic group profiling. While pointing out various applications of taxonomy adaption, we suggest some future work that can take advantage of burgeoning Web 2.0 services for online targeted marketing, counterterrorism in connecting dots, and community tracking.
The synthetic dataset includes two source domains (D1 and D2) and one target domain with conflicting conditional probabilities.
We consider the characterization of muscle fatigue through noninvasive sensing mechanism such as surface electromyography (SEMG). While changes in the properties of SEMG signals with respect to muscle fatigue have been reported in the literature, the large variation in these signals across different individuals makes the task of modeling and classification of SEMG signals challenging. Indeed, the variation in SEMG parameters from subject to subject creates differences in the data distribution. In this paper, we propose a transfer learning framework based on the multi-source domain adaptation methodology for detecting different stages of fatigue using SEMG signals, that addresses the distribution differences. In the proposed framework, the SEMG data of a subject represent a domain; data from multiple subjects in the training set form the multiple source domains and the test subject data form the target domain. SEMG signals are predominantly different in conditional probability distribution across subjects. The key feature of the proposed framework is a novel weighting scheme that addresses the conditional probability distribution differences across multiple domains (subjects). We have validated the proposed framework on Surface Electromyogram signals collected from 8 people during a fatigue-causing repetitive gripping activity. Comprehensive experiments on the SEMG data set demonstrate that the proposed method improves the classification accuracy by 20% to 30% over the cases without any domain adaptation method and by 13% to 30% over the existing state-of-the-art domain adaptation methods.
Microarray technology has generated enormous amounts of high-dimensional gene expression data, providing a unique platform for exploring gene regulatory networks. However, the curse of dimensionality plagues effort to analyze these high throughput data. Linear Discriminant Analysis (LDA) and Biased Discriminant Analysis (BDA) are two popular techniques for dimension reduction, which pay attention to different roles of the positive and negative samples in finding discriminating subspace. However, the drawbacks of these two methods are obvious: LDA has limited efficiency in classifying sample data from subclasses with different distributions, and BDA does not account for the underlying distribution of negative samples. In this paper, we propose a novel dimension reduction technique for microarray analysis: Adaptive Discriminant Analysis (ADA), which effectively exploits favorable attributes of both BDA and LDA and avoids their unfavorable ones. ADA can find a good discriminative subspace with adaptation to different sample distributions. It not only alleviates the problem of high dimensionality, but also enhances the classification performance in the subspace with naïve Bayes classifier. To learn the best model fitting the real scenario, boosted Adaptive Discriminant Analysis is further proposed. Extensive experiments on the yeast cell cycle regulation data set, and the expression data of the red blood cell cycle in malaria parasite Plasmodium falciparum demonstrate the superior performance of ADA and boosted ADA. We also present some putative genes of specific functional classes predicted by boosted ADA. Their potential functionality is confirmed by independent predictions based on Gene Ontology, demonstrating that ADA and boosted ADA are effective dimension reduction methods for microarray-based classification.
Given a spatial dataset placed on an n ×n grid, our goal is to find the rectangular regions within which subsets of the dataset exhibit anomalous behavior. We develop algorithms that, given any user-supplied arbitrary likelihood function, conduct a likelihood ratio hypothesis test (LRT) over each rectangular region in the grid, rank all of the rectangles based on the computed LRT statistics, and return the top few most interesting rectangles. To speed this process, we develop methods to prune rectangles without computing their associated LRT statistics.
A low-rank approximation to a matrix A is a matrix with significantly smaller rank than A, and which is close to A according to some norm. Many practical applications involving the use of large matrices focus on low-rank approximations. By reducing the rank or dimensionality of the data, we reduce the complexity of analyzing the data. The singular value decomposition is the most popular low-rank matrix approximation. However, due to its expensive computational requirements, it has often been considered intractable for practical applications involving massive data. Recent developments have tried to address this problem, with several methods proposed to approximate the decomposition with better asymptotic runtime. We present an empirical study of these techniques on a variety of dense and sparse datasets. We find that a sampling approach of Drineas, Kannan and Mahoney is often, but not always, the best performing method. This method gives solutions with high accuracy much faster than classical SVD algorithms, on large sparse datasets in particular. Other modern methods, such as a recent algorithm by Rokhlin and Tygert, also offer savings compared to classical SVD algorithms. The older sampling methods of Achlioptas and McSherry are shown to sometimes take longer than classical SVD.
Studying the association between quantitative phenotype (such as height or weight) and single nucleotide polymorphisms (SNPs) is an important problem in biology. To understand underlying mechanisms of complex phenotypes, it is often necessary to consider joint genetic effects across multiple SNPs. ANOVA (analysis of variance) test is routinely used in association study. Important findings from studying gene-gene (SNP-pair) interactions are appearing in the literature. However, the number of SNPs can be up to millions. Evaluating joint effects of SNPs is a challenging task even for SNP-pairs. Moreover, with large number of SNPs correlated, permutation procedure is preferred over simple Bonferroni correction for properly controlling family-wise error rate and retaining mapping power, which dramatically increases the computational cost of association study. In this article, we study the problem of finding SNP-pairs that have significant associations with a given quantitative phenotype. We propose an efficient algorithm, FastANOVA, for performing ANOVA tests on SNP-pairs in a batch mode, which also supports large permutation test. We derive an upper bound of SNP-pair ANOVA test, which can be expressed as the sum of two terms. The first term is based on single-SNP ANOVA test. The second term is based on the SNPs and independent of any phenotype permutation. Furthermore, SNP-pairs can be organized into groups, each of which shares a common upper bound. This allows for maximum reuse of intermediate computation, efficient upper bound estimation, and effective SNP-pair pruning. Consequently, FastANOVA only needs to perform the ANOVA test on a small number of candidate SNP-pairs without the risk of missing any significant ones. Extensive experiments demonstrate that FastANOVA is orders of magnitude faster than the brute-force implementation of ANOVA tests on all SNP pairs. The principles used in FastANOVA can be applied to categorical phenotypes and other statistics such as Chi-square test.
In this article, we study the problem of approximate local triangle counting in large graphs. Namely, given a large graph G =( V,E ) we want to estimate as accurately as possible the number of triangles incident to every node v ∈ V in the graph. We consider the question both for undirected and directed graphs. The problem of computing the global number of triangles in a graph has been considered before, but to our knowledge this is the first contribution that addresses the problem of approximate local triangle counting with a focus on the efficiency issues arising in massive graphs and that also considers the directed case. The distribution of the local number of triangles and the related local clustering coefficient can be used in many interesting applications. For example, we show that the measures we compute can help detect the presence of spamming activity in large-scale Web graphs, as well as to provide useful features for content quality assessment in social networks. For computing the local number of triangles (undirected and directed), we propose two approximation algorithms, which are based on the idea of min-wise independent permutations [Broder et al. 1998]. Our algorithms operate in a semi-streaming fashion, using O (| V |) space in main memory and performing O (log | V |) sequential scans over the edges of the graph. The first algorithm we describe in this article also uses O (| E |) space of external memory during computation, while the second algorithm uses only main memory. We present the theoretical analysis as well as experimental results on large graphs, demonstrating the practical efficiency of our approach.
In this article we propose a moment-based method for studying models and model selection mea- sures. By focusing on the probabilistic space of classifiers induced by the classification algorithm rather than on that of datasets, we obtain efficient characterizations for computing the moments, which is followed by visualization of the resulting formulae that are too complicated for direct in- terpretation. By assuming the data to be drawn independently and identically distributed from the underlying probability distribution, and by going over the space of all possible datasets, we establish general relationships between the generalization error, hold-out-set error, cross-validation error, and leave-one-out error. We later exemplify the method and the results by studying the behavior of the errors for the naive Bayes classifier.
Two Gaussian distributions with means µ 1 = 0 and µ 2 = 2 and very different variances. Cluster 1 is surrounded by cluster 2. K-means would assign the point x = −3 to cluster 1, even though the posterior probability of cluster 2 is higher.
Average WAE over 10 runs evaluated on all or only on the foreground clusters. The error bars are plus or minus one standard deviation. Lower WAE is better.
Illustration of rolling clustering with cluster B disappearing and cluster C forming in the 1993-1997 time period. By using the common documents from the overlapping time periods, we can match the clusters and track their evolution. 
How clusters merge or split. Cluster 6 was labeled "Constrained Satisfaction" and cluster 15 "Optimization". Cluster 9 was "Internet Traffic Management" and 17 "Distributed Computing".
Insight into the growth (or shrinkage) of “knowledge communities” of authors that build on each other's work can be gained by studying the evolution over time of clusters of documents. We cluster documents based on the documents they cite in common using the Streemer clustering method, which finds cohesive foreground clusters (the knowledge communities) embedded in a diffuse background. We build predictive models with features based on the citation structure, the vocabulary of the papers, and the affiliations and prestige of the authors and use these models to study the drivers of community growth and the predictors of how widely a paper will be cited. We find that scientific knowledge communities tend to grow more rapidly if their publications build on diverse information and use narrow vocabulary and that papers that lie on the periphery of a community have the highest impact, while those not in any community have the lowest impact.
We discover communities from social network data and analyze the community evolution. These communities are inherent characteristics of human interaction in online social networks, as well as paper citation networks. Also, communities may evolve over time, due to changes to individuals' roles and social status in the network as well as changes to individuals' research interests. We present an innovative algorithm that deviates from the traditional two-step approach to analyze community evolutions. In the traditional approach, communities are first detected for each time slice, and then compared to determine correspondences. We argue that this approach is inappropriate in applications with noisy data. In this paper, we propose FacetNet for analyzing communities and their evolutions through a robust unified process. This novel framework will discover communities and capture their evolution with temporal smoothness given by historic community structures. Our approach relies on formulating the problem in terms of maximum a posteriori (MAP) estimation, where the community structure is estimated both by the observed networked data and by the prior distribution given by historic community structures. Then we develop an iterative algorithm, with proven low time complexity, which is guaranteed to converge to an optimal solution. We perform extensive experimental studies, on both synthetic datasets and real datasets, to demonstrate that our method discovers meaningful communities and provides additional insights not directly obtainable from traditional methods.
Using frequent patterns to analyze data has been one of the fundamental approaches in many data mining applications. Research in frequent pattern mining has so far mostly focused on developing efficient algorithms to discover various kinds of frequent patterns, but little attention has been paid to the important next step—interpreting the discovered frequent patterns. Although the compression and summarization of frequent patterns has been studied in some recent work, the proposed techniques there can only annotate a frequent pattern with nonsemantical information (e.g., support), which provides only limited help for a user to understand the patterns. In this article, we study the novel problem of generating semantic annotations for frequent patterns. The goal is to discover the hidden meanings of a frequent pattern by annotating it with in-depth, concise, and structured information. We propose a general approach to generate such an annotation for a frequent pattern by constructing its context model, selecting informative context indicators, and extracting representative transactions and semantically similar patterns. This general approach can well incorporate the user's prior knowledge, and has potentially many applications, such as generating a dictionary-like description for a pattern, finding synonym patterns, discovering semantic relations, and summarizing semantic classes of a set of frequent patterns. Experiments on different datasets show that our approach is effective in generating semantic pattern annotations.
Publishing data about individuals without revealing sensitive information about them is an important problem. In recent years, a new definition of privacy called k-anonymity has gained popularity. In a k-anonymized dataset, each record is indistinguishable from at least k − 1 other records with respect to certain identifying attributes. In this article, we show using two simple attacks that a k-anonymized dataset has some subtle but severe privacy problems. First, an attacker can discover the values of sensitive attributes when there is little diversity in those sensitive attributes. This is a known problem. Second, attackers often have background knowledge, and we show that k-anonymity does not guarantee privacy against attackers using background knowledge. We give a detailed analysis of these two attacks, and we propose a novel and powerful privacy criterion called ℓ-diversity that can defend against such attacks. In addition to building a formal foundation for ℓ-diversity, we show in an experimental evaluation that ℓ-diversity is practical and can be implemented efficiently.
The TIPS data structure.
b) depicts the discernibility ratio D R with adversary's knowledge L = 2, 4, 6, anonymity threshold 20 ≤ K ≤ 100, and a fixed confidence threshold C = 20%. D R generally increases as K increases, so it exhibits some tradeoff between data privacy and data utility. As L increases, D R increases rapidly because more generalization is required to ensure each equivalence group has at least K records. To illustrate the benefit of our proposed L KC-privacy model over the traditional K-anonymity model, we measure the discernibility ratio, denoted D R TradK , on traditional K-anonymous solutions produced by the TDR method in Fung et al. [2007]. D R TradK − D R, representing the benefit of our model, spans from 0.1 to 0.45. This indicates a significant improvement on data quality by making a reasonable assumption on limiting the adversary's knowledge within L known values. Note, the solutions produced by TDR do not prevent attribute linkages although they have higher discernibility ratio. Figure 6(a) depicts the classification error CE with adversary's knowledge L = 2, 4, 6, anonymity threshold 20 ≤ K ≤ 100, and confidence threshold C = 20% on the Adult dataset. BE = 14.7% and U E = 24.5%. For L = 2, CE − BE is less than 1% and U E − CE spans from 8.9% to 9.5%. For L = 4 and L = 6, CE − BE spans from 1.1% to 4.1%, and U E − CE spans from 5.8% to 8.8%.
Adult dataset.
Scalability (L = 4, K = 20, C = 100%).
Sharing healthcare data has become a vital requirement in healthcare system management; however, inappropriate sharing and usage of healthcare data could threaten patients’ privacy. In this article, we study the privacy concerns of sharing patient information between the Hong Kong Red Cross Blood Transfusion Service (BTS) and the public hospitals. We generalize their information and privacy requirements to the problems of centralized anonymization and distributed anonymization, and identify the major challenges that make traditional data anonymization methods not applicable. Furthermore, we propose a new privacy model called LKC-privacy to overcome the challenges and present two anonymization algorithms to achieve LKC-privacy in both the centralized and the distributed scenarios. Experiments on real-life data demonstrate that our anonymization algorithms can effectively retain the essential information in anonymous data for data analysis and is scalable for anonymizing large datasets.
Decision makers of companies often face the dilemma of whether to release data for knowledge discovery, vis a vis the risk of disclosing proprietary or sensitive information. Among the var- ious methods employed for "sanitizing" the data prior to disclosure, we focus in this paper on anonymization, given its widespread use in practice. We do due diligence to the question "just how safe is the anonymized data". We consider both the scenarios when the hacker has no infor- mation, and more realistically, when the hacker may have partial information about items in the domain. We conduct our analyses in the context of frequent set mining and address the safety question at two di!erent levels: (i) how likely are the ident ities of individual items cracked (i.e. reidentified by the hacker), and (ii) how likely are sets of items cracked. For capturing the prior knowledge of the hacker, we propose a belief function, which amounts to an educated guess of the frequency of each item. For various classes of belief functions, which correspond to di!erent degrees of prior knowledge, we derive formulas for computing the expected number of cracks of single items and for itemsets, the probability of cracking the itemsets. While obtaining the exact values for the more general situations is computationally hard, we propose a series of heuristics called the O-estimates. They are easy to compute, and are shown to be fairly accurate, justified by empirical results on real benchmark datasets. Based on the O-estimates, we propose a recipe for the decision makers to resolve their dilemma. Our recipe operates at two di!erent levels, depending on whether the data owner wants to reason in terms of single items or sets of items (or both). Finally, we present techniques using which a hacker's knowledge of correlation in terms of co-occurrence of items can be incorporated into our framework of disclosure risk analysis and present experimental results demonstrating how this knowledge a!ects the heuristic estimates we have developed.
Question Answering Communities such as Naver, Baidu Knows, and Yahoo! Answers have emerged as popular, and often effective, means of information seeking on the web. By posting questions for other participants to answer, information seekers can obtain specific answers to their questions. Users of CQA portals have already contributed millions of questions, and received hundreds of millions of answers from other participants. However, CQA is not always effective: in some cases, a user may obtain a perfect answer within minutes, and in others it may require hours—and sometimes days—until a satisfactory answer is contributed. We investigate the problem of predicting information seeker satisfaction in collaborative question answering communities, where we attempt to predict whether a question author will be satisfied with the answers submitted by the community participants. We present a general prediction model, and develop a variety of content, structure, and community-focused features for this task. Our experimental results, obtained from a large-scale evaluation over thousands of real questions and user ratings, demonstrate the feasibility of modeling and predicting asker satisfaction. We complement our results with a thorough investigation of the interactions and information seeking patterns in question answering communities that correlate with information seeker satisfaction. We also explore personalized models of asker satisfaction, and show that when sufficient interaction history exists, personalization can significantly improve prediction accuracy over a “one-size-fits-all” model. Our models and predictions could be useful for a variety of applications, such as user intent inference, answer ranking, interface design, and query suggestion and routing.
In this article, we define a knowledge-sharing community in a question-answering forum as a set of askers and authoritative users such that, within each community, askers exhibit more homogeneous behavior in terms of their interactions with authoritative users than elsewhere. A procedure for discovering members of such a community is devised. As a case study, we focus on Yahoo! Answers, a large and diverse online question-answering service. Our contribution is twofold. First, we propose a method for automatic identification of authoritative actors in Yahoo! Answers. To this end, we estimate and then model the authority scores of participants as a mixture of gamma distributions. The number of components in the mixture is determined using the Bayesian Information Criterion (BIC), while the parameters of each component are estimated using the Expectation-Maximization (EM) algorithm. This method allows us to automatically discriminate between authoritative and nonauthoritative users. Second, we represent the forum environment as a type of transactional data such that each transaction summarizes the interaction of an asker with a specific set of authoritative users. Then, to group askers on the basis of their interactions with authoritative users, we propose a parameter-free transaction data clustering algorithm which is based on a novel criterion function. The identified clusters correspond to the communities that we aim to discover. To evaluate the suitability of our clustering algorithm, we conduct a series of experiments on both synthetic data and public real-life data. Finally, we put our approach to work using data from Yahoo! Answers which represent users’ activities over one full year.
Record linkage is an important data integration task that has many practical uses for matching, merging and duplicate removal in large and diverse databases. However, quadratic scalability for the brute force approach of comparing all possible pairs of records necessitates the design of appropriate indexing or blocking techniques. The aim of these techniques is to cheaply remove candidate record pairs that are unlikely to match. We design and evaluate an efficient and highly scalable blocking approach based on suffix arrays. Our suffix grouping technique exploits the ordering used by the index to merge similar blocks at marginal extra cost, resulting in a much higher accuracy while retaining the high scalability of the base suffix array method. Efficiently grouping similar suffixes is carried out with the use of a sliding window technique. We carry out an in-depth analysis of our method and show results from experiments using real and synthetic data, which highlight the importance of using efficient indexing and blocking in real-world applications where datasets contain millions of records. We extend our disk-based methods with the capability to utilise main memory based storage to construct Bloom filters, which we have found to cause significant speedup by reducing the number of costly database queries by up to 70% in real data. We give practical implementation details and show how Bloom filters can be easily applied to Suffix Array based indexing.
Finding information is becoming a major part of our daily life. Entire sectors, from Web users to scientists and intelligence analysts, are increasingly struggling to keep up with the larger and larger amounts of content published every day. With this much data, it is often easy to miss the big picture. In this article, we investigate methods for automatically connecting the dots---providing a structured, easy way to navigate within a new topic and discover hidden connections. We focus on the news domain: given two news articles, our system automatically finds a coherent chain linking them together. For example, it can recover the chain of events starting with the decline of home prices (January 2007), and ending with the health care debate (2009). We formalize the characteristics of a good chain and provide a fast search-driven algorithm to connect two fixed endpoints. We incorporate user feedback into our framework, allowing the stories to be refined and personalized. We also provide a method to handle partially-specified endpoints, for users who do not know both ends of a story. Finally, we evaluate our algorithm over real news data. Our user studies demonstrate that the objective we propose captures the users’ intuitive notion of coherence, and that our algorithm effectively helps users understand the news.
A swap in a 0-1 matrix. 
Convergence: x-axis: the number of steps (× the number of 1's in the data); y-axis: the number of frequent itemsets in the sampled datasets, divided by the number of frequent itemsets in the original dataset. 
The problem of assessing the signicance of data mining re- sults on high-dimensional 0{1 data sets has been studied extensively in the literature. For problems such as mining frequent sets and nding correlations, signicance testing can be done by, e.g., chi-square tests, or many other meth- ods. However, the results of such tests depend only on the specic attributes and not on the dataset as a whole. More- over, the tests are more dicult to apply to sets of patterns or other complex results of data mining. In this paper, we consider a simple randomization technique that deals with this shortcoming. The approach consists of producing ran- dom datasets that have the same row and column margins with the given dataset, computing the results of interest on the randomized instances, and comparing them against the results on the actual data. This randomization technique can be used to assess the results of many dieren t types of data mining algorithms, such as frequent sets, clustering, and rankings. To generate random datasets with given mar- gins, we use variations of a Markov chain approach, which is based on a simple swap operation. We give theoretical results on the eciency of dieren t randomization methods, and apply the swap randomization method to several well- known datasets. Our results indicate that for some datasets the structure discovered by the data mining algorithms is a random artifact, while for other datasets the discovered structure conveys meaningful information.
We investigate the general model of mining associations in a temporal database, where the exhibition periods of items are allowed to be different from one to another. The database is divided into partitions according to the time granularity imposed. Such temporal association rules allow us to observe short-term but interesting patterns that are absent when the whole range of the database is evaluated altogether. Prior work may omit some temporal association rules and thus have limited practicability. To remedy this and to give more precise frequent exhibition periods of frequent temporal itemsets, we devise an efficient algorithm Twain (standing for TWo end AssocIation miNer.) Twain not only generates frequent patterns with more precise frequent exhibition periods, but also discovers more interesting frequent patterns. Twain employs Start time and End time of each item to provide precise frequent exhibition period while progressively handling itemsets from one partition to another. Along with one scan of the database, Twain can generate frequent 2-itemsets directly according to the cumulative filtering threshold. Then, Twain adopts the scan reduction technique to generate all frequent k-itemsets (k > 2) from the generated frequent 2-itemsets. Theoretical properties of Twain are derived as well in this article. The experimental results show that Twain outperforms the prior works in the quality of frequent patterns, execution time, I/O cost, CPU overhead and scalability.
Plot of evaluation time against number of itemsets evaluated 
Self-sufficient itemsets are those whose frequency cannot explained solely by the frequency of either their subsets or of their supersets. We argue that itemsets that are not self-sufficient will often be of little interest to the data analyst, as their frequency should be expected once that of the itemsets on which their frequency depends is known. We present statistical tests for statistically sound discovery of self-sufficient itemsets, and computational techniques that allow those tests to be applied as a post-processing step for any itemset discovery algorithm. We also present a measure for assessing the degree of potential interest in an itemset that complements these statistical measures.
Clustering real-time stream data is an important and challenging problem. Existing algorithms such as CluStream are based on the k -means algorithm. These clustering algorithms have difficulties finding clusters of arbitrary shapes and handling outliers. Further, they require the knowledge of k and user-specified time window. To address these issues, this article proposes D-Stream , a framework for clustering stream data using a density-based approach. Our algorithm uses an online component that maps each input data record into a grid and an offline component that computes the grid density and clusters the grids based on the density. The algorithm adopts a density decaying technique to capture the dynamic changes of a data stream and a attraction-based mechanism to accurately generate cluster boundaries. Exploiting the intricate relationships among the decay factor, attraction, data density, and cluster structure, our algorithm can efficiently and effectively generate and adjust the clusters in real time. Further, a theoretically sound technique is developed to detect and remove sporadic grids mapped by outliers in order to dramatically improve the space and time efficiency of the system. The technique makes high-speed data stream clustering feasible without degrading the clustering quality. The experimental results show that our algorithm has superior quality and efficiency, can find clusters of arbitrary shapes, and can accurately recognize the evolving behaviors of real-time data streams.
Top-cited authors
Muthuramakrishnan Venkitasubramaniam
Jure Leskovec
  • Stanford University
Zhi-Hua Zhou
  • Nanjing University
Peer Kröger
  • Ludwig-Maximilians-University of Munich
Fei Tony Liu
  • Monash University (Australia)