Showing papers presented at "Knowledge Discovery and Data Mining in 2009"

Efficient influence maximization in social networks

Abstract: Influence maximization is the problem of finding a small subset of nodes (seed nodes) in a social network that could maximize the spread of influence. In this paper, we study the efficient influence maximization from two complementary directions. One is to improve the original greedy algorithm of [5] and its improvement [7] to further reduce its running time, and the second is to propose new degree discount heuristics that improves influence spread. We evaluate our algorithms by experiments on two large academic collaboration graphs obtained from the online archival database arXiv.org. Our experimental results show that (a) our improved greedy algorithm achieves better running time comparing with the improvement of [7] with matching influence spread, (b) our degree discount heuristics achieve much better influence spread than classic degree and centrality-based heuristics, and when tuned for a specific influence cascade model, it achieves almost matching influence thread with the greedy algorithm, and more importantly (c) the degree discount heuristics run only in milliseconds while even the improved greedy algorithms run in hours in our experiment graphs with a few tens of thousands of nodes.Based on our results, we believe that fine-tuned heuristics may provide truly scalable solutions to the influence maximization problem with satisfying influence spread and blazingly fast running time. Therefore, contrary to what implied by the conclusion of [5] that traditional heuristics are outperformed by the greedy approximation algorithm, our results shed new lights on the research of heuristic algorithms.

Collaborative filtering with temporal dynamics

Abstract: Customer preferences for products are drifting over time. Product perception and popularity are constantly changing as new selection emerges. Similarly, customer inclinations are evolving, leading them to ever redefine their taste. Thus, modeling temporal dynamics should be a key when designing recommender systems or general customer preference models. However, this raises unique challenges. Within the eco-system intersecting multiple products and customers, many different characteristics are shifting simultaneously, while many of them influence each other and often those shifts are delicate and associated with a few data instances. This distinguishes the problem from concept drift explorations, where mostly a single concept is tracked. Classical time-window or instance-decay approaches cannot work, as they lose too much signal when discarding data instances. A more sensitive approach is required, which can make better distinctions between transient effects and long term patterns. The paradigm we offer is creating a model tracking the time changing behavior throughout the life span of the data. This allows us to exploit the relevant components of all data instances, while discarding only what is modeled as being irrelevant. Accordingly, we revamp two leading collaborative filtering recommendation approaches. Evaluation is made on a large movie rating dataset by Netflix. Results are encouraging and better than those previously reported on this dataset.

Meme-tracking and the dynamics of the news cycle

Abstract: Tracking new topics, ideas, and "memes" across the Web has been an issue of considerable interest. Recent work has developed methods for tracking topic shifts over long time scales, as well as abrupt spikes in the appearance of particular named entities. However, these approaches are less well suited to the identification of content that spreads widely and then fades over time scales on the order of days - the time scale at which we perceive news and events.We develop a framework for tracking short, distinctive phrases that travel relatively intact through on-line text; developing scalable algorithms for clustering textual variants of such phrases, we identify a broad class of memes that exhibit wide spread and rich variation on a daily basis. As our principal domain of study, we show how such a meme-tracking approach can provide a coherent representation of the news cycle - the daily rhythms in the news media that have long been the subject of qualitative interpretation but have never been captured accurately enough to permit actual quantitative analysis. We tracked 1.6 million mainstream media sites and blogs over a period of three months with the total of 90 million articles and we find a set of novel and persistent temporal patterns in the news cycle. In particular, we observe a typical lag of 2.5 hours between the peaks of attention to a phrase in the news media and in blogs respectively, with divergent behavior around the overall peak and a "heartbeat"-like pattern in the handoff between news and blogs. We also develop and analyze a mathematical model for the kinds of temporal variation that the system exhibits.

Social influence analysis in large-scale networks

Abstract: In large social networks, nodes (users, entities) are influenced by others for various reasons. For example, the colleagues have strong influence on one's work, while the friends have strong influence on one's daily life. How to differentiate the social influences from different angles(topics)? How to quantify the strength of those social influences? How to estimate the model on real large networks?To address these fundamental questions, we propose Topical Affinity Propagation (TAP) to model the topic-level social influence on large networks. In particular, TAP can take results of any topic modeling and the existing network structure to perform topic-level influence propagation. With the help of the influence analysis, we present several important applications on real data sets such as 1) what are the representative nodes on a given topic? 2) how to identify the social influences of neighboring nodes on a particular node?To scale to real large networks, TAP is designed with efficient distributed learning algorithms that is implemented and tested under the Map-Reduce framework. We further present the common characteristics of distributed learning algorithms for Map-Reduce. Finally, we demonstrate the effectiveness and efficiency of TAP on real large data sets.

Time series shapelets: a new primitive for data mining

Abstract: Classification of time series has been attracting great interest over the past decade. Recent empirical evidence has strongly suggested that the simple nearest neighbor algorithm is very difficult to beat for most time series problems. While this may be considered good news, given the simplicity of implementing the nearest neighbor algorithm, there are some negative consequences of this. First, the nearest neighbor algorithm requires storing and searching the entire dataset, resulting in a time and space complexity that limits its applicability, especially on resource-limited sensors. Second, beyond mere classification accuracy, we often wish to gain some insight into the data.In this work we introduce a new time series primitive, time series shapelets, which addresses these limitations. Informally, shapelets are time series subsequences which are in some sense maximally representative of a class. As we shall show with extensive empirical evaluations in diverse domains, algorithms based on the time series shapelet primitives can be interpretable, more accurate and significantly faster than state-of-the-art classifiers.

TrustWalker: a random walk model for combining trust-based and item-based recommendation

Abstract: Collaborative filtering is the most popular approach to build recommender systems and has been successfully employed in many applications. However, it cannot make recommendations for so-called cold start users that have rated only a very small number of items. In addition, these methods do not know how confident they are in their recommendations. Trust-based recommendation methods assume the additional knowledge of a trust network among users and can better deal with cold start users, since users only need to be simply connected to the trust network. On the other hand, the sparsity of the user item ratings forces the trust-based approach to consider ratings of indirect neighbors that are only weakly trusted, which may decrease its precision. In order to find a good trade-off, we propose a random walk model combining the trust-based and the collaborative filtering approach for recommendation. The random walk model allows us to define and to measure the confidence of a recommendation. We performed an evaluation on the Epinions dataset and compared our model with existing trust-based and collaborative filtering methods.

Financial incentives and the "performance of crowds"

Abstract: The relationship between financial incentives and performance, long of interest to social scientists, has gained new relevance with the advent of web-based "crowd-sourcing" models of production. Here we investigate the effect of compensation on performance in the context of two experiments, conducted on Amazon's Mechanical Turk (AMT). We find that increased financial incentives increase the quantity, but not the quality, of work performed by participants, where the difference appears to be due to an "anchoring" effect: workers who were paid more also perceived the value of their work to be greater, and thus were no more motivated than workers paid less. In contrast with compensation levels, we find the details of the compensation scheme do matter---specifically, a "quota" system results in better work for less pay than an equivalent "piece rate" system. Although counterintuitive, these findings are consistent with previous laboratory studies, and may have real-world analogs as well.

Beyond blacklists: learning to detect malicious web sites from suspicious URLs

Abstract: Malicious Web sites are a cornerstone of Internet criminal activities. As a result, there has been broad interest in developing systems to prevent the end user from visiting such sites. In this paper, we describe an approach to this problem based on automated URL classification, using statistical methods to discover the tell-tale lexical and host-based properties of malicious Web site URLs. These methods are able to learn highly predictive models by extracting and automatically analyzing tens of thousands of features potentially indicative of suspicious URLs. The resulting classifiers obtain 95-99% accuracy, detecting large numbers of malicious Web sites from their URLs, with only modest false positives.

Differentially private recommender systems: Building privacy into the Netflix Prize contenders

Abstract: We consider the problem of producing recommendations from collective user behavior while simultaneously providing guarantees of privacy for these users. Specifically, we consider the Netflix Prize data set, and its leading algorithms, adapted to the framework of differential privacy.Unlike prior privacy work concerned with cryptographically securing the computation of recommendations, differential privacy constrains a computation in a way that precludes any inference about the underlying records from its output. Such algorithms necessarily introduce uncertainty--i.e., noise--to computations, trading accuracy for privacy.We find that several of the leading approaches in the Netflix Prize competition can be adapted to provide differential privacy, without significantly degrading their accuracy. To adapt these algorithms, we explicitly factor them into two parts, an aggregation/learning phase that can be performed with differential privacy guarantees, and an individual recommendation phase that uses the learned correlations and an individual's data to provide personalized recommendations. The adaptations are non-trivial, and involve both careful analysis of the per-record sensitivity of the algorithms to calibrate noise, as well as new post-processing steps to mitigate the impact of this noise.We measure the empirical trade-off between accuracy and privacy in these adaptations, and find that we can provide non-trivial formal privacy guarantees while still outperforming the Cinematch baseline Netflix provides.

Relational learning via latent social dimensions

Abstract: Social media such as blogs, Facebook, Flickr, etc., presents data in a network format rather than classical IID distribution. To address the interdependency among data instances, relational learning has been proposed, and collective inference based on network connectivity is adopted for prediction. However, connections in social media are often multi-dimensional. An actor can connect to another actor for different reasons, e.g., alumni, colleagues, living in the same city, sharing similar interests, etc. Collective inference normally does not differentiate these connections. In this work, we propose to extract latent social dimensions based on network information, and then utilize them as features for discriminative learning. These social dimensions describe diverse affiliations of actors hidden in the network, and the discriminative learning can automatically determine which affiliations are better aligned with the class labels. Such a scheme is preferred when multiple diverse relations are associated with the same network. We conduct extensive experiments on social media data (one from a real-world blog site and the other from a popular content sharing site). Our model outperforms representative relational learning methods based on collective inference, especially when few labeled data are available. The sensitivity of this model and its connection to existing methods are also examined.

Safe-Level-SMOTE: Safe-Level-Synthetic Minority Over-Sampling TEchnique for Handling the Class Imbalanced Problem

Abstract: The class imbalanced problem occurs in various disciplines when one of target classes has a tiny number of instances comparing to other classes. A typical classifier normally ignores or neglects to detect a minority class due to the small number of class instances. SMOTE is one of over-sampling techniques that remedies this situation. It generates minority instances within the overlapping regions. However, SMOTE randomly synthesizes the minority instances along a line joining a minority instance and its selected nearest neighbours, ignoring nearby majority instances. Our technique called Safe-Level-SMOTE carefully samples minority instances along the same line with different weight degree, called safe level. The safe level computes by using nearest neighbour minority instances. By synthesizing the minority instances more around larger safe level, we achieve a better accuracy performance than SMOTE and Borderline-SMOTE.

Finding a team of experts in social networks

Abstract: Given a task T, a pool of individuals X with different skills, and a social network G that captures the compatibility among these individuals, we study the problem of finding X, a subset of X, to perform the task. We call this the TEAM FORMATION problem. We require that members of X' not only meet the skill requirements of the task, but can also work effectively together as a team. We measure effectiveness using the communication cost incurred by the subgraph in G that only involves X'. We study two variants of the problem for two different communication-cost functions, and show that both variants are NP-hard. We explore their connections with existing combinatorial problems and give novel algorithms for their solution. To the best of our knowledge, this is the first work to consider the TEAM FORMATION problem in the presence of a social network of individuals. Experiments on the DBLP dataset show that our framework works well in practice and gives useful and intuitive results.

WhereNext: a location predictor on trajectory pattern mining

Abstract: The pervasiveness of mobile devices and location based services is leading to an increasing volume of mobility data.This side eect provides the opportunity for innovative methods that analyse the behaviors of movements. In this paper we propose WhereNext, which is a method aimed at predicting with a certain level of accuracy the next location of a moving object. The prediction uses previously extracted movement patterns named Trajectory Patterns, which are a concise representation of behaviors of moving objects as sequences of regions frequently visited with a typical travel time. A decision tree, named T-pattern Tree, is built and evaluated with a formal training and test process. The tree is learned from the Trajectory Patterns that hold a certain area and it may be used as a predictor of the next location of a new trajectory finding the best matching path in the tree. Three dierent best matching methods to classify a new moving object are proposed and their impact on the quality of prediction is studied extensively. Using Trajectory Patterns as predictive rules has the following implications: (I) the learning depends on the movement of all available objects in a certain area instead of on the individual history of an object; (II) the prediction tree intrinsically contains the spatio-temporal properties that have emerged from the data and this allows us to define matching methods that striclty depend on the properties of such movements. In addition, we propose a set of other measures, that evaluate a priori the predictive power of a set of Trajectory Patterns. This measures were tuned on a real life case study. Finally, an exhaustive set of experiments and results on the real dataset are presented.

New ensemble methods for evolving data streams

Abstract: Advanced analysis of data streams is quickly becoming a key area of data mining research as the number of applications demanding such processing increases. Online mining when such data streams evolve over time, that is when concepts drift or change completely, is becoming one of the core issues. When tackling non-stationary concepts, ensembles of classifiers have several advantages over single classifier methods: they are easy to scale and parallelize, they can adapt to change quickly by pruning under-performing parts of the ensemble, and they therefore usually also generate more accurate concept descriptions. This paper proposes a new experimental data stream framework for studying concept drift, and two new variants of Bagging: ADWIN Bagging and Adaptive-Size Hoeffding Tree (ASHT) Bagging. Using the new experimental framework, an evaluation study on synthetic and real-world datasets comprising up to ten million examples shows that the new ensemble methods perform very well compared to several known methods.

Regression-based latent factor models

Abstract: We propose a novel latent factor model to accurately predict response for large scale dyadic data in the presence of features. Our approach is based on a model that predicts response as a multiplicative function of row and column latent factors that are estimated through separate regressions on known row and column features. In fact, our model provides a single unified framework to address both cold and warm start scenarios that are commonplace in practical applications like recommender systems, online advertising, web search, etc. We provide scalable and accurate model fitting methods based on Iterated Conditional Mode and Monte Carlo EM algorithms. We show our model induces a stochastic process on the dyadic space with kernel (covariance) given by a polynomial function of features. Methods that generalize our procedure to estimate factors in an online fashion for dynamic applications are also considered. Our method is illustrated on benchmark datasets and a novel content recommendation application that arises in the context of Yahoo! Front Page. We report significant improvements over several commonly used methods on all datasets.

Ranking-based clustering of heterogeneous information networks with star network schema

Abstract: A heterogeneous information network is an information networkcomposed of multiple types of objects. Clustering on such a network may lead to better understanding of both hidden structures of the network and the individual role played by every object in each cluster. However, although clustering on homogeneous networks has been studied over decades, clustering on heterogeneous networks has not been addressed until recently.A recent study proposed a new algorithm, RankClus, for clustering on bi-typed heterogeneous networks. However, a real-world network may consist of more than two types, and the interactions among multi-typed objects play a key role at disclosing the rich semantics that a network carries. In this paper, we study clustering of multi-typed heterogeneous networks with a star network schema and propose a novel algorithm, NetClus, that utilizes links across multityped objects to generate high-quality net-clusters. An iterative enhancement method is developed that leads to effective ranking-based clustering in such heterogeneous networks. Our experiments on DBLP data show that NetClus generates more accurate clustering results than the baseline topic model algorithm PLSA and the recently proposed algorithm, RankClus. Further, NetClus generates informative clusters, presenting good ranking and cluster membership information for each attribute object in each net-cluster.

Sentiment analysis of blogs by combining lexical knowledge with text classification

Abstract: The explosion of user-generated content on the Web has led to new opportunities and significant challenges for companies, that are increasingly concerned about monitoring the discussion around their products. Tracking such discussion on weblogs, provides useful insight on how to improve products or market them more effectively. An important component of such analysis is to characterize the sentiment expressed in blogs about specific brands and products. Sentiment Analysis focuses on this task of automatically identifying whether a piece of text expresses a positive or negative opinion about the subject matter. Most previous work in this area uses prior lexical knowledge in terms of the sentiment-polarity of words. In contrast, some recent approaches treat the task as a text classification problem, where they learn to classify sentiment based only on labeled training data. In this paper, we present a unified framework in which one can use background lexical information in terms of word-class associations, and refine this information for specific domains using any available training examples. Empirical results on diverse domains show that our approach performs better than using background knowledge or training data in isolation, as well as alternative approaches to using lexical knowledge with text classification.

Fast approximate spectral clustering

Abstract: Spectral clustering refers to a flexible class of clustering procedures that can produce high-quality clusterings on small data sets but which has limited applicability to large-scale problems due to its computational complexity of O(n3) in general, with n the number of data points. We extend the range of spectral clustering by developing a general framework for fast approximate spectral clustering in which a distortion-minimizing local transformation is first applied to the data. This framework is based on a theoretical analysis that provides a statistical characterization of the effect of local distortion on the mis-clustering rate. We develop two concrete instances of our general framework, one based on local k-means clustering (KASP) and one based on random projection trees (RASP). Extensive experiments show that these algorithms can achieve significant speedups with little degradation in clustering accuracy. Specifically, our algorithms outperform k-means by a large margin in terms of accuracy, and run several times faster than approximate spectral clustering based on the Nystrom method, with comparable accuracy and significantly smaller memory footprint. Remarkably, our algorithms make it possible for a single machine to spectral cluster data sets with a million observations within several minutes.

Collective annotation of Wikipedia entities in web text

Abstract: To take the first step beyond keyword-based search toward entity-based search, suitable token spans ("spots") on documents must be identified as references to real-world entities from an entity catalog. Several systems have been proposed to link spots on Web pages to entities in Wikipedia. They are largely based on local compatibility between the text around the spot and textual metadata associated with the entity. Two recent systems exploit inter-label dependencies, but in limited ways. We propose a general collective disambiguation approach. Our premise is that coherent documents refer to entities from one or a few related topics or domains. We give formulations for the trade-off between local spot-to-entity compatibility and measures of global coherence between entities. Optimizing the overall entity assignment is NP-hard. We investigate practical solutions based on local hill-climbing, rounding integer linear programs, and pre-clustering entities followed by local optimization within clusters. In experiments involving over a hundred manually-annotated Web pages and tens of thousands of spots, our approaches significantly outperform recently-proposed algorithms.

Efficient methods for topic model inference on streaming document collections

Abstract: Topic models provide a powerful tool for analyzing large text collections by representing high dimensional data in a low dimensional subspace. Fitting a topic model given a set of training documents requires approximate inference techniques that are computationally expensive. With today's large-scale, constantly expanding document collections, it is useful to be able to infer topic distributions for new documents without retraining the model. In this paper, we empirically evaluate the performance of several methods for topic inference in previously unseen documents, including methods based on Gibbs sampling, variational inference, and a new method inspired by text classification. The classification-based inference method produces results similar to iterative inference methods, but requires only a single matrix multiplication. In addition to these inference methods, we present SparseLDA, an algorithm and data structure for evaluating Gibbs sampling distributions. Empirical results indicate that SparseLDA can be approximately 20 times faster than traditional LDA and provide twice the speedup of previously published fast sampling methods, while also using substantially less memory.

Learning optimal ranking with tensor factorization for tag recommendation

Abstract: Tag recommendation is the task of predicting a personalized list of tags for a user given an item. This is important for many websites with tagging capabilities like last.fm or delicious. In this paper, we propose a method for tag recommendation based on tensor factorization (TF). In contrast to other TF methods like higher order singular value decomposition (HOSVD), our method RTF ('ranking with tensor factorization') directly optimizes the factorization model for the best personalized ranking. RTF handles missing values and learns from pairwise ranking constraints. Our optimization criterion for TF is motivated by a detailed analysis of the problem and of interpretation schemes for the observed data in tagging systems. In all, RTF directly optimizes for the actual problem using a correct interpretation of the data. We provide a gradient descent algorithm to solve our optimization problem. We also provide an improved learning and prediction method with runtime complexity analysis for RTF. The prediction runtime of RTF is independent of the number of observations and only depends on the factorization dimensions. Besides the theoretical analysis, we empirically show that our method outperforms other state-of-the-art tag recommendation methods like FolkRank, PageRank and HOSVD both in quality and prediction runtime.

Combining link and content for community detection: a discriminative approach

Abstract: In this paper, we consider the problem of combining link and content analysis for community detection from networked data, such as paper citation networks and Word Wide Web. Most existing approaches combine link and content information by a generative model that generates both links and contents via a shared set of community memberships. These generative models have some shortcomings in that they failed to consider additional factors that could affect the community memberships and isolate the contents that are irrelevant to community memberships. To explicitly address these shortcomings, we propose a discriminative model for combining the link and content analysis for community detection. First, we propose a conditional model for link analysis and in the model, we introduce hidden variables to explicitly model the popularity of nodes. Second, to alleviate the impact of irrelevant content attributes, we develop a discriminative model for content analysis. These two models are unified seamlessly via the community memberships. We present efficient algorithms to solve the related optimization problems based on bound optimization and alternating projection. Extensive experiments with benchmark data sets show that the proposed framework significantly outperforms the state-of-the-art approaches for combining link and content analysis for community detection.

On compressing social networks

Abstract: Motivated by structural properties of the Web graph that support efficient data structures for in memory adjacency queries, we study the extent to which a large network can be compressed. Boldi and Vigna (WWW 2004), showed that Web graphs can be compressed down to three bits of storage per edge; we study the compressibility of social networks where again adjacency queries are a fundamental primitive. To this end, we propose simple combinatorial formulations that encapsulate efficient compressibility of graphs. We show that some of the problems are NP-hard yet admit effective heuristics, some of which can exploit properties of social networks such as link reciprocity. Our extensive experiments show that social networks and the Web graph exhibit vastly different compressibility characteristics.

DOULION: counting triangles in massive graphs with a coin

Abstract: Counting the number of triangles in a graph is a beautiful algorithmic problem which has gained importance over the last years due to its significant role in complex network analysis. Metrics frequently computed such as the clustering coefficient and the transitivity ratio involve the execution of a triangle counting algorithm. Furthermore, several interesting graph mining applications rely on computing the number of triangles in the graph of interest.In this paper, we focus on the problem of counting triangles in a graph. We propose a practical method, out of which all triangle counting algorithms can potentially benefit. Using a straightforward triangle counting algorithm as a black box, we performed 166 experiments on real-world networks and on synthetic datasets as well, where we show that our method works with high accuracy, typically more than 99% and gives significant speedups, resulting in even ≈ 130 times faster performance.

A New Local Distance-Based Outlier Detection Approach for Scattered Real-World Data

Abstract: Detecting outliers which are grossly different from or inconsistent with the remaining dataset is a major challenge in real-world KDD applications. Existing outlier detection methods are ineffective on scattered real-world datasets due to implicit data patterns and parameter setting issues. We define a novel Local Distance-based Outlier Factor (LDOF) to measure the outlier-ness of objects in scattered datasets which addresses these issues. LDOF uses the relative location of an object to its neighbours to determine the degree to which the object deviates from its neighbourhood. We present theoretical bounds on LDOF's false-detection probability. Experimentally, LDOF compares favorably to classical KNN and LOF based outlier detection. In particular it is less sensitive to parameter values.

Issues in evaluation of stream learning algorithms

Abstract: Learning from data streams is a research area of increasing importance. Nowadays, several stream learning algorithms have been developed. Most of them learn decision models that continuously evolve over time, run in resource-aware environments, detect and react to changes in the environment generating data. One important issue, not yet conveniently addressed, is the design of experimental work to evaluate and compare decision models that evolve over time. There are no golden standards for assessing performance in non-stationary environments. This paper proposes a general framework for assessing predictive stream learning algorithms. We defend the use of Predictive Sequential methods for error estimate - the prequential error. The prequential error allows us to monitor the evolution of the performance of models that evolve over time. Nevertheless, it is known to be a pessimistic estimator in comparison to holdout estimates. To obtain more reliable estimators we need some forgetting mechanism. Two viable alternatives are: sliding windows and fading factors. We observe that the prequential error converges to an holdout estimator when estimated over a sliding window or using fading factors. We present illustrative examples of the use of prequential error estimators, using fading factors, for the tasks of: i) assessing performance of a learning algorithm; ii) comparing learning algorithms; iii) hypothesis testing using McNemar test; and iv) change detection using Page-Hinkley test. In these tasks, the prequential error estimated using fading factors provide reliable estimators. In comparison to sliding windows, fading factors are faster and memory-less, a requirement for streaming applications. This paper is a contribution to a discussion in the good-practices on performance assessment when learning dynamic models that evolve over time.

On the tradeoff between privacy and utility in data publishing

Abstract: In data publishing, anonymization techniques such as generalization and bucketization have been designed to provide privacy protection. In the meanwhile, they reduce the utility of the data. It is important to consider the tradeoff between privacy and utility. In a paper that appeared in KDD 2008, Brickell and Shmatikov proposed an evaluation methodology by comparing privacy gain with utility gain resulted from anonymizing the data, and concluded that "even modest privacy gains require almost complete destruction of the data-mining utility". This conclusion seems to undermine existing work on data anonymization. In this paper, we analyze the fundamental characteristics of privacy and utility, and show that it is inappropriate to directly compare privacy with utility. We then observe that the privacy-utility tradeoff in data publishing is similar to the risk-return tradeoff in financial investment, and propose an integrated framework for considering privacy-utility tradeoff, borrowing concepts from the Modern Portfolio Theory for financial investment. Finally, we evaluate our methodology on the Adult dataset from the UCI machine learning repository. Our results clarify several common misconceptions about data utility and provide data publishers useful guidelines on choosing the right tradeoff between privacy and utility.

Exploiting Wikipedia as external knowledge for document clustering

Abstract: In traditional text clustering methods, documents are represented as "bags of words" without considering the semantic information of each document. For instance, if two documents use different collections of core words to represent the same topic, they may be falsely assigned to different clusters due to the lack of shared core words, although the core words they use are probably synonyms or semantically associated in other forms. The most common way to solve this problem is to enrich document representation with the background knowledge in an ontology. There are two major issues for this approach: (1) the coverage of the ontology is limited, even for WordNet or Mesh, (2) using ontology terms as replacement or additional features may cause information loss, or introduce noise. In this paper, we present a novel text clustering method to address these two issues by enriching document representation with Wikipedia concept and category information. We develop two approaches, exact match and relatedness-match, to map text documents to Wikipedia concepts, and further to Wikipedia categories. Then the text documents are clustered based on a similarity metric which combines document content information, concept information as well as category information. The experimental results using the proposed clustering framework on three datasets (20-newsgroup, TDT2, and LA Times) show that clustering performance improves significantly by enriching document representation with Wikipedia concepts and categories.

Frequent pattern mining with uncertain data

Abstract: This paper studies the problem of frequent pattern mining with uncertain data. We will show how broad classes of algorithms can be extended to the uncertain data setting. In particular, we will study candidate generate-and-test algorithms, hyper-structure algorithms and pattern growth based algorithms. One of our insightful observations is that the experimental behavior of different classes of algorithms is very different in the uncertain case as compared to the deterministic case. In particular, the hyper-structure and the candidate generate-and-test algorithms perform much better than tree-based algorithms. This counter-intuitive behavior is an important observation from the perspective of algorithm design of the uncertain variation of the problem. We will test the approach on a number of real and synthetic data sets, and show the effectiveness of two of our approaches over competitive techniques.

OpinionMiner: a novel machine learning system for web opinion mining and extraction

Abstract: Merchants selling products on the Web often ask their customers to share their opinions and hands-on experiences on products they have purchased. Unfortunately, reading through all customer reviews is difficult, especially for popular items, the number of reviews can be up to hundreds or even thousands. This makes it difficult for a potential customer to read them to make an informed decision. The OpinionMiner system designed in this work aims to mine customer reviews of a product and extract high detailed product entities on which reviewers express their opinions. Opinion expressions are identified and opinion orientations for each recognized product entity are classified as positive or negative. Different from previous approaches that employed rule-based or statistical techniques, we propose a novel machine learning approach built under the framework of lexicalized HMMs. The approach naturally integrates multiple important linguistic features into automatic learning. In this paper, we describe the architecture and main components of the system. The evaluation of the proposed method is presented based on processing the online product reviews from Amazon and other publicly available datasets.