Projecting multi-dimensional data to a lower-dimensional visual display is a commonly used approach for identifying and analyzing patterns in data. Many dimensionality reduction techniques exist for generating visual ...
详细信息
Projecting multi-dimensional data to a lower-dimensional visual display is a commonly used approach for identifying and analyzing patterns in data. Many dimensionality reduction techniques exist for generating visual embeddings, but it is often hard to avoid cluttered projections when the data is large in size and noisy. For many application users who are not machine learning experts, it is difficult to control the process in order to improve the "readability" of the projection and at the same time to understand their quality. In this paper, we propose a simple interactive feature transformation approach that allows the analyst to de-clutter the visualization by gradually transforming the original feature space based on existing class knowledge. By changing a single parameter, the user can easily decide the desired trade-off between structural preservation and the visual quality during the transforming process. The proposed approach integrates semi-interactive feature transformation techniques as well as a variety of quality measures to help analysts generate uncluttered projections and understand their quality. (C) 2014 Elsevier B.V. All rights reserved.
Dimensionality Reduction is a commonly used method to reduce the number of dimensions of data. In this work, we verified its influence in classification process using combinations of projection techniques as dimension...
详细信息
ISBN:
(纸本)9781538680230
Dimensionality Reduction is a commonly used method to reduce the number of dimensions of data. In this work, we verified its influence in classification process using combinations of projection techniques as dimensionality reduction algorithms. We also used Naive Bayes and SMO as classifiers.
Dimensionality reduction is commonly applied to multidimensionaldata to reduce the complexity of their analysis. In visual analysis systems, projections embed multidimensionaldata into 2D or 3D spaces for graphical ...
详细信息
Dimensionality reduction is commonly applied to multidimensionaldata to reduce the complexity of their analysis. In visual analysis systems, projections embed multidimensionaldata into 2D or 3D spaces for graphical representation. To facilitate a robust and accurate analysis, essential characteristics of the multidimensionaldata shall be preserved when projecting. Orthographic star coordinates is a state-of-the-art linear projection method that avoids distortion of multidimensional clusters by restricting interactive exploration to orthographic projections. However, existing numerical methods for computing orthographic star coordinates have a number of limitations when putting them into practice. We overcome these limitations by proposing the novel concept of shape preserving star coordinates where shape preservation is assured using a superset of orthographic projections. Our scheme is explicit, exact, simple, fast, parameter-free, and stable. To maintain a valid shape-preserving star-coordinates configuration during user interaction with one of the star-coordinates axes, we derive an algorithm that only requires us to modify the configuration of one additional compensatory axis. Different design goals can be targeted by using different strategies for selecting the compensatory axis. We propose and discuss four strategies including a strategy that approximates orthographic star coordinates very well and a data-driven strategy. We further present shape-preserving morphing strategies between two shape-preserving configurations, which can be adapted for the generation of data tours. We apply our concept to multiple data analysis scenarios to document its applicability and validate its desired properties.
data mining models are frequently used to represent and summarize meaningful properties in data. However such models are usually not suitable for interactive data exploration and visualization. This paper proposes the...
详细信息
ISBN:
(纸本)9781467389426
data mining models are frequently used to represent and summarize meaningful properties in data. However such models are usually not suitable for interactive data exploration and visualization. This paper proposes the use of multidimensionalprojection together with the Ubiquitous Self-Organizing Map algorithm (UbiSOM), a novel variant of the well-known self-organizing map algorithm that was specifically tuned for stream datamining. The resulting high-dimensional projection system is then studied for interactive data analysis and visualization. A prototype was developed where, at each moment, the user can visualize the information from different perspectives. Direct interaction with the system during stream processing is possible both by changing the projection, by optimizing the projection view for maximizing variance or by filtering the incoming data series. Experiments in two distinct datasets show the importance and relevance of conjoining multidimensional data projection with UbiSOM.
Teleconnections refer to links between regions that are distant to each other, but nevertheless exhibit some relation. The study of such teleconnections is a well-known task in climate research. Climate simulation sha...
详细信息
Teleconnections refer to links between regions that are distant to each other, but nevertheless exhibit some relation. The study of such teleconnections is a well-known task in climate research. Climate simulation shall model known teleconnections. Detecting teleconnections in climate simulations is a crucial aspect in judging the quality of the simulation output. It is common practice to run scripts to execute a sequence of analysis steps on the climate simulations to search for teleconnections. Such a scripting approach is not flexible and targeted towards one specific goal. It is desirable to have one tool that allows for a flexible analysis of all teleconnection patterns with a dataset. We present such a tool, where the extracted information is provided in an intuitive visual form to users, who then can interactively explore the data. We developed an analysis workflow that is modeled around four views showing different facets of the data with coordinated interaction. We present a teleconnection study with simulation ensembles and reanalysis data obtained by data assimilation to observe how well the teleconnectivity patterns match and to demonstrate the effectiveness of our tool.
暂无评论