Danilo Medeiros Eler

Piecewise laplacian-based projection for interactive data exploration and organization

Multidimensional projection has emerged as an important visualization tool in applications involving the visual analysis of high‐dimensional data. However, high precision projection methods are either computationally expensive or not flexible enough to enable feedback from user interaction into the projection process. A built‐in mechanism that dynamically adapts the projection based on direct user intervention would make the technique more useful for a larger range of applications and data sets. In this paper we propose the Piecewise Laplacian‐based Projection (PLP), a novel multidimensional projection technique, that, due to the local nature of its formulation, enables a versatile mechanism to interact with projected data and to allow interactive changes to alter the projection map dynamically, a capability unique of this technique. We exploit the flexibility provided by PLP in two interactive projection‐based applications, one designed to organize pictures visually and another to build music playlists. These applications illustrate the usefulness of PLP in handling high‐dimensional data in a flexible and highly visual way. We also compare PLP with the currently most promising projections in terms of precision and speed, showing that it performs very well also according to these quality criteria.

Visual analysis of image collections

Multidimensional Visualization techniques are invaluable tools for analysis of structured and unstructured data with variable dimensionality. This paper introduces PEx-Image—Projection Explorer for Images—axa0tool aimed at supporting analysis of image collections. The tool supports a methodology that employs interactive visualizations to aid user-driven feature detection and classification tasks, thus offering improved analysis and exploration capabilities. The visual mappings employ similarity-based multidimensional projections and point placement to layout the data on a plane for visual exploration. In addition to its application to image databases, we also illustrate how the proposed approach can be successfully employed in simultaneous analysis of different data types, such as text and images, offering a common visual representation for data expressed in different modalities.

Characterizing 3D shapes using fractal dimension

Developments in techniques for modeling and digitizing have made the use of 3D models popular to a large number of new applications. With the diffusion and spreading of 3D models employment, the demand for efficient search and retrieval methods is high. Researchers have dedicated effort to investigate and overcome the problem of 3D shape retrieval. In this work, we propose a new way to employ shape complexity analysis methods, such as the fractal dimension, to perform the 3D shape characterization for those purposes. This approach is described and experimental results are performed on a 3D models data set. We also compare the technique to two other known methods for 3D model description, reported in literature, namely shape histograms and shape distributions. The technique presented here has performed considerably better than any of the others in the experiments.

Topic-Based Coordination for Visual Analysis of Evolving Document Collections

Document interpretation is a crucial task in many visual analytics applications, made harder by the widespread availability of freely available textual files. In this paper we propose an approach based on topic detection coupled with multiple coordinated views to assist analysis of time varying document collections. Given multiple document maps built from a set of text files, we define a strategy to support users locating the evolution of topics addressed by the documents, along various time steps. The approach is supported by a new algorithm for topic extraction from texts, also introduced. Finally, we show several examples illustrating how the proposed strategy may be applied in the analysis of document collections.

Multidimensional Visualization to Support Analysis of Image Collections

Multidimensional visualization techniques are invaluable tools for analysis of structured and unstructured data with variable dimensionality. This paper introduces a methodology and a software tool called PEx-Image - Projection Explorer for Image for analysis and exploration of image collections employing visualizations. The visual mappings proposed here are similarity-based multidimensional projections and point placements, which layout the data on a plane for visual exploration. The proposed approach supports various image analysis tasks such as feature selection and classification, improving data exploration capabilities. We also illustrate how it can be successfully employed in simultaneous analysis of different data types, such as text and images, offering a common visual representation for data expressed in different modalities.

Using Otsu's Threshold Selection Method for Eliminating Terms in Vector Space Model Computation

Visualization techniques have proved to be valuable tools to support textual data exploration. Dimensionality reduction techniques have been widely used to produce visual representation of document collections. Focusing on multidimensional projection techniques, good visual results are produced depending on how representative terms to discriminate the documents are chosen to compose the vector space model (VSM). To define a good VSM it is necessary to apply filters during the preprocessing in order to eliminate terms using their frequency. For that, the user must evaluate the term frequency histogram based on his/her expertise in the text subject and decide the threshold value for frequency cut. Usually it is a trial and error approach that requires the user to verify the quality of visual representation after each trial. In this paper, we propose an automatic approach that applies the Otsus Threshold Selection Method for computing a threshold using a term frequency histogram. We conducted experiments that have shown our approach generates visual representations as good as those generated with a threshold obtained by trial and error approach. The contribution of our approach is that users with non expertise are able to generate good visual representations and the time to get a good threshold is decreased.

Visual Data Exploration to Feature Space Definition

Many image-related applications rely on the fact that the dataset under investigation is correctly represented by features. However, defining a set of features that properly represents a dataset is still a challenging and, in most cases, an exhausting task. Most of the available techniques, especially when a large number of features is considered, are based on purely quantitative statistical measures or approaches based on artificial intelligence, and normally are “black-boxes” to the user. The approach proposed here seeks to open this “black-box” by means of visual representations, enabling users to get insight about the meaning and representativeness of the features computed from different feature extraction algorithms and sets of parameters. The results show that, as the combination of sets of features and changes in parameters improves the quality of the visual representation, the accuracy of the classification for the computed features also improves. The results strongly suggest that our approach can be successfully employed as a guidance to defining and understanding a set of features that properly represents an image dataset.

Silhouette-based feature selection for classification of medical images

Classification is an important task for computer-aided diagnosis systems (CADs). However, many classifiers may not perform well, presenting poor generalization and high computational cost, especially when dealing with high-dimensional datasets. Thus, feature selection can greatly mitigate these problems. In this paper, we propose two filter-based feature selection algorithms that calculate the simplified silhouette statistic as evaluation function: the silhouette-based greedy search (SiGS) and the silhouette-based genetic algorithm search (SiGAS). Silhouette statistic is used to guide the search for features that provide better class separability. Experiments performed on three datasets have shown that the SiGAS algorithm overcomes traditional filter algorithms, such as CFS, FCBF and reliefF. It also outperforms a similar algorithm, kNNGAS, based on genetic algorithm that minimizes the classification error of k-nearest neighbors. Additionally, results have shown that SiGAS produces better accuracy than SiGS.

Analysis of Document Pre-Processing Effects in Text and Opinion Mining

Typically, textual information is available as unstructured data, which require processing so that data mining algorithms can handle such data; this processing is known as the pre-processing step in the overall text mining process. This paper aims at analyzing the strong impact that the pre-processing step has on most mining tasks. Therefore, we propose a methodology to vary distinct combinations of pre-processing steps and to analyze which pre-processing combination allows high precision. In order to show different combinations of pre-processing methods, experiments were performed by comparing some combinations such as stemming, term weighting, term elimination based on low frequency cut and stop words elimination. These combinations were applied in text and opinion mining tasks, from which correct classification rates were computed to highlight the strong impact of the pre-processing combinations. Additionally, we provide graphical representations from each pre-processing combination to show how visual approaches are useful to show the processing effects on document similarities and group formation (i.e., cohesion and separation).

Hadoop Cluster Deployment: A Methodological Approach

For a long time, data has been treated as a general problem because it just represents fractions of an event without any relevant purpose. However, the last decade has been just about information and how to get it. Seeking meaning in data and trying to solve scalability problems, many frameworks have been developed to improve data storage and its analysis. As a framework, Hadoop was presented as a powerful tool to deal with large amounts of data. However, it still causes doubts about how to deal with its deployment and if there is any reliable method to compare the performance of distinct Hadoop clusters. This paper presents a methodology based on benchmark analysis to guide the Hadoop cluster deployment. The experiments employed The Apache Hadoop and the Hadoop distributions of Cloudera, Hortonworks, and MapR, analyzing the architectures on local and on clouding—using centralized and geographically distributed servers. The results show the methodology can be dynamically applied on a reliable comparison among different architectures. Additionally, the study suggests that the knowledge acquired can be used to improve the data analysis process by understanding the Hadoop architecture.