of a dissertation at the University of Miami. Dissertation supervised by Professor Mei-Ling Shyu. No. of pages in text. (153) The development in information science has enabled an explosive growth of data, which attracts more and more researchers to engage in the field of big data analytics. Noticeably, in many real-world applications, large amounts of data are imbalanced data since the events of interests occur infrequently. Classification of imbalanced data is an important research problem as lots of real-world datasets have skewed class distributions in which the majority of instances (examples) belong to one class and far fewer instances belong to the others. A classifier induced from an imbalanced dataset is more likely to be biased towards the majority classes and shows very poor classification accuracy on the minority classes. While in many applications, the minority instances actually represent the concept of interest (e.g., fraud in banking operations, abnormal cell in medical data, etc.), and the detection of these rare events has become more important. Despite extensive research efforts, rare event mining remains one of the most challenging problems in information retrieval, especially for multimedia big data. To tackle this challenge, in this dissertation, we propose an extended deep learning approach to achieve promising performance in classifying largely skewed multimedia dataset. Specifically, we investigate the integration of bootstrapping methods and a state-of-the-art deep learning approach, Convolutional Neural Networks (CNNs), with extensive empirical studies. Considering the fact that deep learning approaches such as CNNs are usually computationally expensive, we propose to feed low-level features to CNNs and prove its feasibility in achieving promising performance while saving a lot of training time. Furthermore, since big training datasets are required to train CNNs, we propose to extract features from pre-trained CNN models and feed those features to another full connected neural network. Implementations in big data environments show promising performance of our model in handling big datasets with respect to feasibility and scalability. In order to further improve the classification results and bridge the semantic gap between high-level concepts and low-level visual features, correlation discovery in semantic concept mining is worth exploring. Though inter-concept correlations have been utilized to address this issue recently, the very small number of instances in the minority classes often lead to the detection of imprecise correlations and unsatisfactory classification results. Meanwhile, correlation discovery is a computationally intensive task in the sense that it requires a deep analysis of very large and growing repositories. This dissertation further proposes a novel concept correlation analysis strategy framework that utilizes the correlations between the retrieval scores and labels. By integrating the correlation information, the proposed framework can help imbalanced data classification and enhance rare class (event or concept) mining even with trivial scores from the minority classes. Not only deep learning but also numerous other classification algorithms have been developed for a variety of data types. However, it is nearly impossible for one classifier to perform the best in all kinds of datasets all the time. Therefore, ensemble learning models which aim to take advantages of different classifiers have received a lot of attentions recently. In this dissertation, a scalable classifier ensemble framework assisted by a set of “judgers” is also proposed to integrate the outputs from multiple classifiers for multimedia big data classification. Specifically, based on the confusion matrices of different classifiers, a set of judgers are organized into a hierarchically structured decision model. A testing instance is first input to different classifiers, and then the classification results are passed to the proposed hierarchical structured decision model to derive the final result. The ensemble system can be run on Spark, which is designed for big data processing. All the proposed components are evaluated on multimedia datasets containing different kinds of data. The experimental results show the effectiveness of our framework in classifying severely imbalanced data with promising performance, and demonstrate that the proposed classifier ensemble framework outperforms several state-of-the-art model fusion approaches. Furthermore, the proposed framework is applied to two realworld applications, i.e., deep learning based text data analysis on an Amazon review dataset and efficient large-scale stance Analysis in Twitter, and achieves promising results in both. In additional, we also design a web-based information retrieval system and identify several future directions that could be explored to further improve the current work. To my dear parents and grandparents

by broad-scale models, then stacking independent species distribution models to predict species assemblages (sensu Guisan and Rahbek 2011, Calabrese et al. 2014) will provide misleading predictions of fine-scale community assembly. Thus, a better understanding of species associations across scales could improve predictions of the dynamics of local community composition in changing environments. The goal of this paper is to improve the tools needed to detect interspecific associations from co-occurrence data. We first briefly describe the development of co-occurrence methods and then draw from different lines of research to present a more complete and flexible general framework for inferring species associations that overcomes multiple challenges faced by previous approaches.

Within the scope of this article we went further into the question to what extent particularistic attributes - social origin and gender - can affect selection processes (1) in access to and (2) in later career attainment after achieving the doctoral degree. The analyses are based on a questionnaire survey (n = 2 244) among doctoral degree holders achieving the doctoral degree in six selected disciplines (biology, electrical engineering, German studies, mathematics, social sciences, and business studies/ economics) at German universities. In terms of our first object of investigation, the analyses show that in four out of six disciplines doctoral degree holders are a selected group compared to university graduates with regard to both social origin and gender. In terms of our second object of investigation - the impact of particularistic attributes on several indicators of further career attainment after achieving the doctoral degree (career inside or outside higher education and science, career position and income) - the results point to a stronger impact of gender compared to social origin.

With the extensive use of smart devices and blooming popularity of social media websites such as Flickr, YouTube, Twitter, and Facebook, we have witnessed an explosion of multimedia data. The amount of data nowadays is formidable without effective big data technologies. It is well-acknowledged that multimedia high-level semantic concept mining and retrieval has become an important research topic; while the semantic gap (i.e., the gap between the low-level features and high-level concepts) makes it even more challenging. To address these challenges, it requires the joint research efforts from both big data mining and multimedia areas. In particular, the correlations among the classes can provide important context cues to help bridge the semantic gap. However, correlation discovery is computationally expensive due to the huge amount of data. In this paper, a novel multimedia big data mining system based on the MapReduce framework is proposed to discover negative correlations for semantic concept mining and retrieval. Furthermore, the proposed multimedia big data mining system consists of a big data processing platform with Mesos for efficient resource management and with Cassandra for handling data across multiple data centers. Experimental results on the TRECVID benchmark datasets demonstrate the feasibility and the effectiveness of the proposed multimedia big data mining system with negative correlation discovery for semantic concept mining and retrieval.

William Kruskal (Bill) was a distinguished statistician who spent virtually his entire professional career at the University of Chicago, and who had a lasting impact on the Institute of Mathematical Statistics and on the field of statistics more broadly, as well as on many who came in contact with him. Bill passed away last April following an extended illness, and on May 19, 2005, the University of Chicago held a memorial service at which several of Bill's colleagues and collaborators spoke along with members of his family and other friends. This biography and the accompanying commentaries derive in part from brief presentations on that occasion, along with recollections and input from several others. Bill was known personally to most of an older generation of statisticians as an editor and as an intellectual and professional leader. In 1994, Statistical Science published an interview by Sandy Zabell (Vol. 9, 285--303) in which Bill looked back on selected events in his professional life. One of the purposes of the present biography and accompanying commentaries is to reintroduce him to old friends and to introduce him for the first time to new generations of statisticians who never had an opportunity to interact with him and to fall under his influence.

We show that, when the three-way association level among the three binary variables, X, U 1 and U 2 is fixed, D P e pr(X e 1¦U 1 e 1) - pr(X e 1¦U 1 e 0) increases as the cross-product ratio of U 1 and U 2 increases under the assumption that X is positively associated with U 1 and U 2 . We then discuss some implications of this property. Copyright 2007, Oxford University Press.

We propose Fuzzy UTASTAR, a method for inferring fuzzy utility functions from a partial preorder of options evaluated on multiple criteria. It is an extension of the well-known UTASTAR method capable to handle both ordinary (crisp) and fuzzy evaluation data. This property gives much flexibility to decision makers because the majority of real-life decision problems involve a considerable level of uncertainty that hinders them from assigning exact evaluations (scores) to options. In case all evaluation data are crisp the method behaves exactly as the original UTASTAR. The proposed method builds fuzzy additive value functions taking as input a partial preorder on a subset of the options, called reference set, along with their associated scores on the criteria. The resulting fuzzy utility functions can subsequently be used to estimate the (fuzzy) utility of each option, thus allowing their ranking, prioritization, selection or classification. The ranking of the options in partial preorder is as compatible as possible to the original one. The method is implemented into a decision support system and is applied to an example from the transportation domain. Results are found to be in concordance with those of the original method. To the best of our knowledge this is the first attempt to extend UTASTAR method to handle both crisp and fuzzy evaluation data.

We present approaches to estimate content-aware bibliometrics to quantitatively measure the scholarly impact of a publication. Traditional measures to assess quality-related aspects such as citation counts and h-index, do not take into account the content of the publications, which limits their ability to provide rigorous quality-related metrics and can significantly skew the results. Our proposed metric, denoted by Content Informed Index (CII), uses the content of the paper as a source of distant-supervision, to weight the edges of a citation network. These content-aware weights quantify the information in the citation i.e., these weights quantify the extent to which the cited-node informs the citing-node. The weights convert the original unweighted citation network to a weighted one. Consequently, this weighted network can be used to derive impact metrics for the various entities involved, like the publications, authors etc. We evaluate the weights estimated by our approach on three manually annotated datasets, where the annotations quantify the extent of information in the citation. Particularly, we evaluate how well the ranking imposed by our approach associates with the ranking imposed by the manual annotations. The proposed approach achieves up to 103% improvement in performance as compared to second best performing approach.

We present an NLG system that uses Integer Linear Programming to integrate different decisions involved in the generation process. Our approach provides an alternative to pipeline-based sequential processing which has become prevalent in today’s NLG applications.

We present a discrete optimization model based on a linear programming formulation as an alternative to the cascade of classifiers implemented in many language processing systems. Since NLP tasks are correlated with one another, sequential processing does not guarantee optimal solutions. We apply our model in an NLG application and show that it performs better than a pipeline-based system.

We examined relations of the 10 types of values in Schwartz's (1992) theory of voting. Hypotheses were generated by relating the core motivations of each value type to the ideological messages conveyed by party policies and symbols. Eight parties that ran in the 1988 Israeli elections were arrayed by judges on three ideological dimensions: classical liberalism, economic egalitarianism, state and religion. Discriminant analyses yielded a function whose coefficients for value types corresponded to hypotheses for the state and religion dimension and ordered party supporters on this dimension. After dropping religious parties, another value-based function ordered party supporters on the classical liberalism dimension, as predicted. Both functions significantly improved the party classification of voters in a representative national sample (N= 769). Economic egalitarianism, a nonsalient dimension in Israeli politics, was unrelated to values. Results suggest that all types of values may be politically relevant depending on context.

We examine the issue of how to measure the degree to which a functional dependency (FD) is approximate. The primary motivation lies in the fact that approximate FDs represent potentially interesting patterns existent in a table. Their discovery is a valuable data mining problem. However, before algorithms can be developed, a measure must be defined quantifying their approximation degree.First we develop an approximation measure by axiomatizing the following intuition: the degree to which X → Y is approximate in a table T is the degree to which T determines a function from ΠX(T) to ΠY(T). We prove that a unique unnormalized measure satisfies these axioms up to a multiplicative constant. Next we compare the measure developed with two other measures from the literature. In all but one case, we show that the measures can be made to differ as much as possible within normalization. We examine these measure on several real datasets and observe that many of the theoretically possible extreme differences do not bear themselves out. We offer some conclusions as to particular situations where certain measures are more appropriate than others.

We consider properties of the ordinal measure of association defined by the ratio of the proportions of concordant and discordant pairs. For 2 x 2 cross-classification tables, the measure simplifies to the odds ratio. The generalized measure can be used to summarize the difference between two stochastically ordered distributions of an ordinal categorical variable. The ratio of its values for two groups constitutes an odds ratio defined in terms of pairs of observations. Unlike the odds ratio measures proposed by Clayton (1974, Bionxetrika 61, 525-531) for similar purposes, this measure is not linked to specific model assumptions and hence it is more widely applicable as a descriptive measure.

Virtually all AHP studies reported in the literature entail paired comparison ratio judgments from respondents. This paper considers situations where respondents can only supply ordinal judgments with reliability and validity. Results of the paper can also be applied to ratio scaled paired comparison matrices to test for ordinal response inconsistencies. The paper introduces an ordinal Coefficient of Consistence and a nonparametric test of paired comparison response circular triads (inconsistencies).

To determine the relation between breast self‐examination and survival of breast cancer patients, the authors studied 1004 newly diagnosed invasive breast cancer patients in Vermont general hospitals from 1 July 1975 to 31 December 1982. More frequent breast self‐examination was associated with a greater likelihood of the patient's detecting her own cancer, less delay from first symptom to histologic diagnosis, earlier clinical stage, smaller pathologic tumor size, and fewer axillary node metastases. At a median follow‐up time of 52 months (maximum follow‐up, 92 months), 14% of the breast self‐examination performers had died of breast cancer versus 26% of the nonperformers (P < 0.001 based on chi‐square). The product limit survival curve for breast self‐examination performers (N = 424) was significantly better (P < 0.001 by log‐rank test) than for breast self‐examination nonperformers (N = 411). Survival at 5 years was 75% for breast self‐examination performers versus 57% for the nonperformers. The significant survival differences persisted after adjusting for any combination of the covariates age, method of detection, family history of breast cancer, and delay in treatment (P < 0.002). Lead‐time would have to be at least 3 years to negate the apparent beneficial effects of breast self‐examination on survival. The authors conclude that in this population of breast cancer patients, breast self‐examination was related to earlier detection and improved survival.

This paper is concerned with the goodness-of-fit of induced decision trees. Namely, we explore the possibility to measure the goodness-of-fit as it is classically done in statistical modeling. We show how Chi-square statistics and especially the Log-likelihood Ratio statistic that is abundantly used in the modeling of cross tables, can be adapted for induction trees. The Log-likelihood Ratio is well suited for testing the significance of the difference between two nested trees. In addition, we derive from it pseudo R 2’s. We propose also adapted forms of the Akaike (AIC) and Bayesian (BIC) information criteria that prove useful in selecting the best compromise model between fit and complexity.

This paper introduces a wavelet transformation and a cluster ensemble framework using graph theory for clustering gene expression data sets. The experiment results indicate that wavelet transformation and cluster ensemble approaches together yield better clustering results than the single best clustering algorithm on both synthetic and yeast gene expression data sets.

This paper formulates the problem encountered in the first stage of two-stage, binary template matching as a set of hypotheses to be tested, including a hypothesis of ``no object.'' Two new statistics R and G are proposed, based on a likelihood ratio, and are compared to the sum of absolute differences and a correlation measure by analytical approximations and Monte Carlo experiments. Statistical power and a measure of sensitivity to the true location of the object are the criteria. Parameters are the numbers of 1's in object and image, subtemplate size, and parameters reflecting intensity distortion between template and object. One of the proposed statistics R is much more computationally intensive than the other G. Although R is more powerful than G and the other statistics, G is generally more sensitive to the true object location. Statistic G is also more powerful than the sum of absolute differences and correlation. All statistics are robust to incomplete knowledge of distortion parameters. Experiments on Landsat images confirm the sensitivity of G and recommend it for application in the first stage.

This paper discusses the general problem of measuring the association between an independent nominal-scaled variableX and a dependent variableY whose scale of measurement may be interval, ordinal or nominal. The theoretical foundations of a wide range of asymmetric association measures are discussed. Some new measures are also suggested. Fifteen of these association measures, some previously suggested, some new, are singled out for a computer-assisted numerical study in which we compute the value actually taken by each measure under a wide variety of conditions. This comparative study provides important insights into the behavior of the measures.

This paper describes a technique for the rigorous analysis of probabilistic measures of association derived from contingency tables. The technique uses the approach to contingency table analysis first described by Grizzle, Starmer, and Koch and extended by Forthofer and Koch. This approach permits the user to obtain estimates of gamma, Yule's Q, Kendall's taub and tauc, Somer's D, and Goodman and Kruskal's taus, plus estimates of the variances and covariances of those measures of association. With the variance-covariance estimates, the user can examine a wide range of hypotheses for statistical significance.