Document understanding using data mining and machine learning techniques

With the explosive growth of the volume and complexity of document data (e.g., news, blogs, web pages), it has become a necessity to semantically understand documents and deliver meaningful information to users. Areas dealing with these problems are crossing data mining, information retrieval, and machine learning. For example, document clustering and summarization are two fundamental techniques for understanding document data and have attracted much attention in recent years. Given a collection of documents, document clustering aims to partition them into different groups to provide efficient document browsing and navigation mechanisms. One unrevealed area in document clustering is that how to generate meaningful interpretation for the each document cluster resulted from the clustering process. Document summarization is another effective technique for document understanding, which generates a summary by selecting sentences that deliver the major or topic-relevant information in the original documents. How to improve the automatic summarization performance and apply it to newly emerging problems are two valuable research directions. To assist people to capture the semantics of documents effectively and efficiently, the dissertation focuses on developing effective data mining and machine learning algorithms and systems for (1) integrating document clustering and summarization to obtain meaningful document clusters with summarized interpretation, (2) improving document summarization performance and building document understanding systems to solve real-world applications, and (3) summarizing the differences and evolution of multiple document sources.

Exploring the integration of data mining and data visualization

Due to the rapid advances in computing and sensing technologies, enormous amounts of data are being generated everyday in various applications. The integration of data mining and data visualization has been widely used to analyze these massive and complex data sets to discover hidden patterns. For both data mining and visualization to be effective, it is important to include the visualization techniques in the mining process and to generate the discovered patterns for a more comprehensive visual view. In this dissertation, four related problems: dimensionality reduction for visualizing high dimensional datasets, visualization-based clustering evaluation, interactive document mining, and multiple clusterings exploration are studied to explore the integration of data mining and data visualization. In particular, we 1) propose an efficient feature selection method (reliefF + mRMR) for preprocessing high dimensional datasets; 2) present DClusterE to integrate cluster validation with user interaction and provide rich visualization tools for users to examine document clustering results from multiple perspectives; 3) design two interactive document summarization systems to involve users efforts and generate customized summaries from 2D sentence layouts; and 4) propose a new framework which organizes the different input clusterings into a hierarchical tree structure and allows for interactive exploration of multiple clustering solutions.

Hybrid approaches to estimating freeway travel times using point traffic detector data

The accurate and reliable estimation of travel time based on point detector data is needed to support Intelligent Transportation System (ITS) applications. It has been found that the quality of travel time estimation is a function of the method used in the estimation and varies for different traffic conditions. In this study, two hybrid on-line travel time estimation models, and their corresponding off-line methods, were developed to achieve better estimation performance under various traffic conditions, including recurrent congestion and incidents. The first model combines the Mid-Point method, which is a speed-based method, with a traffic flow-based method. The second model integrates two speed-based methods: the Mid-Point method and the Minimum Speed method. In both models, the switch between travel time estimation methods is based on the congestion level and queue status automatically identified by clustering analysis. During incident conditions with rapidly changing queue lengths, shock wave analysis-based refinements are applied for on-line estimation to capture the fast queue propagation and recovery. Travel time estimates obtained from existing speed-based methods, traffic flow-based methods, and the models developed were tested using both simulation and real-world data. The results indicate that all tested methods performed at an acceptable level during periods of low congestion. However, their performances vary with an increase in congestion. Comparisons with other estimation methods also show that the developed hybrid models perform well in all cases. Further comparisons between the on-line and off-line travel time estimation methods reveal that off-line methods perform significantly better only during fast-changing congested conditions, such as during incidents. The impacts of major influential factors on the performance of travel time estimation, including data preprocessing procedures, detector errors, detector spacing, frequency of travel time updates to traveler information devices, travel time link length, and posted travel time range, were investigated in this study. The results show that these factors have more significant impacts on the estimation accuracy and reliability under congested conditions than during uncongested conditions. For the incident conditions, the estimation quality improves with the use of a short rolling period for data smoothing, more accurate detector data, and frequent travel time updates.

The Study of Clustering of Taiwanese Tourists' Motivations to Hong Kong

Abstract Driven by the political and economic forces of cross-strait, Taiwan has become one of the major source markets for Hong Kong tourism industry since 1987. The major purposes of this study were to investigate the following factors (1) The influential factors of travel motivation, (2) The clusters of travel motivations, (3) The marketing segmentation of clusters of Taiwanese tourists to visit Hong Kong. Through ten travel agents, self-report surveys were distributed to collect data from 366 Taiwanese travelers. Hence, four push factors and six pull factors were identified as travel motivations through the factor analysis. Combined with the cluster analysis; five new groups were founded. Finally, five clusters which process unique profiles (location difference, visiting frequency, travel satisfaction, and destination loyalty) were addressed. The suggestions of developing effective market strategies to attract Taiwanese tourists to Hong Kong were also provided.

Mining the online social network data: Influence, summarization, and organization

Online Social Network (OSN) services provided by Internet companies bring people together to chat, share the information, and enjoy the information. Meanwhile, huge amounts of data are generated by those services (they can be regarded as the social media ) every day, every hour, even every minute, and every second. Currently, researchers are interested in analyzing the OSN data, extracting interesting patterns from it, and applying those patterns to real-world applications. However, due to the large-scale property of the OSN data, it is difficult to effectively analyze it. This dissertation focuses on applying data mining and information retrieval techniques to mine two key components in the social media data — users and user-generated contents. Specifically, it aims at addressing three problems related to the social media users and contents: (1) how does one organize the users and the contents? (2) how does one summarize the textual contents so that users do not have to go over every post to capture the general idea? (3) how does one identify the influential users in the social media to benefit other applications, e.g., Marketing Campaign? The contribution of this dissertation is briefly summarized as follows. (1) It provides a comprehensive and versatile data mining framework to analyze the users and user-generated contents from the social media. (2) It designs a hierarchical co-clustering algorithm to organize the users and contents. (3) It proposes multi-document summarization methods to extract core information from the social network contents. (4) It introduces three important dimensions of social influence, and a dynamic influence model for identifying influential users.

Algorithms for Geometric Matching, Clustering, and Covering

With the popularization of GPS-enabled devices such as mobile phones, location data are becoming available at an unprecedented scale. The locations may be collected from many different sources such as vehicles moving around a city, user check-ins in social networks, and geo-tagged micro-blogging photos or messages. Besides the longitude and latitude, each location record may also have a timestamp and additional information such as the name of the location. Time-ordered sequences of these locations form trajectories, which together contain useful high-level information about people's movement patterns.

The first part of this thesis focuses on a few geometric problems motivated by the matching and clustering of trajectories. We first give a new algorithm for computing a matching between a pair of curves under existing models such as dynamic time warping (DTW). The algorithm is more efficient than standard dynamic programming algorithms both theoretically and practically. We then propose a new matching model for trajectories that avoids the drawbacks of existing models. For trajectory clustering, we present an algorithm that computes clusters of subtrajectories, which correspond to common movement patterns. We also consider trajectories of check-ins, and propose a statistical generative model, which identifies check-in clusters as well as the transition patterns between the clusters.

The second part of the thesis considers the problem of covering shortest paths in a road network, motivated by an EV charging station placement problem. More specifically, a subset of vertices in the road network are selected to place charging stations so that every shortest path contains enough charging stations and can be traveled by an EV without draining the battery. We first introduce a general technique for the geometric set cover problem. This technique leads to near-linear-time approximation algorithms, which are the state-of-the-art algorithms for this problem in either running time or approximation ratio. We then use this technique to develop a near-linear-time algorithm for this

shortest-path cover problem.

Mining the Online Social Network Data: Influence, Summarization, and Organization

Online Social Network (OSN) services provided by Internet companies bring people together to chat, share the information, and enjoy the information. Meanwhile, huge amounts of data are generated by those services (they can be regarded as the social media ) every day, every hour, even every minute, and every second. Currently, researchers are interested in analyzing the OSN data, extracting interesting patterns from it, and applying those patterns to real-world applications. However, due to the large-scale property of the OSN data, it is difficult to effectively analyze it. This dissertation focuses on applying data mining and information retrieval techniques to mine two key components in the social media data — users and user-generated contents. Specifically, it aims at addressing three problems related to the social media users and contents: (1) how does one organize the users and the contents? (2) how does one summarize the textual contents so that users do not have to go over every post to capture the general idea? (3) how does one identify the influential users in the social media to benefit other applications, e.g., Marketing Campaign? The contribution of this dissertation is briefly summarized as follows. (1) It provides a comprehensive and versatile data mining framework to analyze the users and user-generated contents from the social media. (2) It designs a hierarchical co-clustering algorithm to organize the users and contents. (3) It proposes multi-document summarization methods to extract core information from the social network contents. (4) It introduces three important dimensions of social influence, and a dynamic influence model for identifying influential users.

Reference list of 265 sources used for the discovery of relationships between data clusters and metadata properties

Visual cluster analysis provides valuable tools that help analysts to understand large data sets in terms of representative clusters and relationships thereof. Often, the found clusters are to be understood in context of belonging categorical, numerical or textual metadata which are given for the data elements. While often not part of the clustering process, such metadata play an important role and need to be considered during the interactive cluster exploration process. Traditionally, linked-views allow to relate (or loosely speaking: correlate) clusters with metadata or other properties of the underlying cluster data. Manually inspecting the distribution of metadata for each cluster in a linked-view approach is tedious, specially for large data sets, where a large search problem arises. Fully interactive search for potentially useful or interesting cluster to metadata relationships may constitute a cumbersome and long process. To remedy this problem, we propose a novel approach for guiding users in discovering interesting relationships between clusters and associated metadata. Its goal is to guide the analyst through the potentially huge search space. We focus in our work on metadata of categorical type, which can be summarized for a cluster in form of a histogram. We start from a given visual cluster representation, and compute certain measures of interestingness defined on the distribution of metadata categories for the clusters. These measures are used to automatically score and rank the clusters for potential interestingness regarding the distribution of categorical metadata. Identified interesting relationships are highlighted in the visual cluster representation for easy inspection by the user. We present a system implementing an encompassing, yet extensible, set of interestingness scores for categorical metadata, which can also be extended to numerical metadata. Appropriate visual representations are provided for showing the visual correlations, as well as the calculated ranking scores. Focusing on clusters of time series data, we test our approach on a large real-world data set of time-oriented scientific research data, demonstrating how specific interesting views are automatically identified, supporting the analyst discovering interesting and visually understandable relationships.

Clustering in non-parametric multivariate analyses.

Non-parametric multivariate analyses of complex ecological datasets are widely used. Following appropriate pre-treatment of the data inter-sample resemblances are calculated using appropriate measures. Ordination and clustering derived from these resemblances are used to visualise relationships among samples (or variables). Hierarchical agglomerative clustering with group-average (UPGMA) linkage is often the clustering method chosen. Using an example dataset of zooplankton densities from the Bristol Channel and Severn Estuary, UK, a range of existing and new clustering methods are applied and the results compared. Although the examples focus on analysis of samples, the methods may also be applied to species analysis. Dendrograms derived by hierarchical clustering are compared using cophenetic correlations, which are also used to determine optimum  in flexible beta clustering. A plot of cophenetic correlation against original dissimilarities reveals that a tree may be a poor representation of the full multivariate information. UNCTREE is an unconstrained binary divisive clustering algorithm in which values of the ANOSIM R statistic are used to determine (binary) splits in the data, to form a dendrogram. A form of flat clustering, k-R clustering, uses a combination of ANOSIM R and Similarity Profiles (SIMPROF) analyses to determine the optimum value of k, the number of groups into which samples should be clustered, and the sample membership of the groups. Robust outcomes from the application of such a range of differing techniques to the same resemblance matrix, as here, result in greater confidence in the validity of a clustering approach.

Clustering in non-parametric multivariate analyses.

Non-parametric multivariate analyses of complex ecological datasets are widely used. Following appropriate pre-treatment of the data inter-sample resemblances are calculated using appropriate measures. Ordination and clustering derived from these resemblances are used to visualise relationships among samples (or variables). Hierarchical agglomerative clustering with group-average (UPGMA) linkage is often the clustering method chosen. Using an example dataset of zooplankton densities from the Bristol Channel and Severn Estuary, UK, a range of existing and new clustering methods are applied and the results compared. Although the examples focus on analysis of samples, the methods may also be applied to species analysis. Dendrograms derived by hierarchical clustering are compared using cophenetic correlations, which are also used to determine optimum  in flexible beta clustering. A plot of cophenetic correlation against original dissimilarities reveals that a tree may be a poor representation of the full multivariate information. UNCTREE is an unconstrained binary divisive clustering algorithm in which values of the ANOSIM R statistic are used to determine (binary) splits in the data, to form a dendrogram. A form of flat clustering, k-R clustering, uses a combination of ANOSIM R and Similarity Profiles (SIMPROF) analyses to determine the optimum value of k, the number of groups into which samples should be clustered, and the sample membership of the groups. Robust outcomes from the application of such a range of differing techniques to the same resemblance matrix, as here, result in greater confidence in the validity of a clustering approach.

Extracting Patterns from Educational Traces via Clustering and Associated Quality Metrics

Clustering algorithms, pattern mining techniques and associated quality metrics emerged as reliable methods for modeling learners’ performance, comprehension and interaction in given educational scenarios. The specificity of available data such as missing values, extreme values or outliers, creates a challenge to extract significant user models from an educational perspective. In this paper we introduce a pattern detection mechanism with-in our data analytics tool based on k-means clustering and on SSE, silhouette, Dunn index and Xi-Beni index quality metrics. Experiments performed on a dataset obtained from our online e-learning platform show that the extracted interaction patterns were representative in classifying learners. Furthermore, the performed monitoring activities created a strong basis for generating automatic feedback to learners in terms of their course participation, while relying on their previous performance. In addition, our analysis introduces automatic triggers that highlight learners who will potentially fail the course, enabling tutors to take timely actions.

MixKMeans: Clustering Question-Answer Archives

Community-driven Question Answering (CQA) systems that crowdsource experiential information in the form of questions and answers and have accumulated valuable reusable knowledge. Clustering of QA datasets from CQA systems provides a means of organizing the content to ease tasks such as manual curation and tagging. In this paper, we present a clustering method that exploits the two-part question-answer structure in QA datasets to improve clustering quality. Our method, {\it MixKMeans}, composes question and answer space similarities in a way that the space on which the match is higher is allowed to dominate. This construction is motivated by our observation that semantic similarity between question-answer data (QAs) could get localized in either space. We empirically evaluate our method on a variety of real-world labeled datasets. Our results indicate that our method significantly outperforms state-of-the-art clustering methods for the task of clustering question-answer archives.

Abnormal Pedestrian Trajectory analysis based on arbitrary-length clustering

This papers examines the use of trajectory distance measures and clustering techniques to define normal
and abnormal trajectories in the context of pedestrian tracking in public spaces. In order to detect abnormal
trajectories, what is meant by a normal trajectory in a given scene is firstly defined. Then every trajectory
that deviates from this normality is classified as abnormal. By combining Dynamic Time Warping and a
modified K-Means algorithms for arbitrary-length data series, we have developed an algorithm for trajectory
clustering and abnormality detection. The final system performs with an overall accuracy of 83% and 75%
when tested in two different standard datasets.

A New Clustering Methodology for the Analysis of Sorted or Categorized Stimuli

This paper introduces a new stochastic clustering methodology devised for the analysis of categorized or sorted data. The methodology reveals consumers' common category knowledge as well as individual differences in using this knowledge for classifying brands in a designated product class. A small study involving the categorization of 28 brands of U.S. automobiles is presented where the results of the proposed methodology are compared with those obtained from KMEANS clustering. Finally, directions for future research are discussed.

Data-Adaptive Modeling using Convex Regression

Thesis (Ph.D.)--University of Washington, 2016-08