检索结果-内蒙古大学图书馆

Human-in-the-loop Extraction of Interpretable Concepts in Deep Learning Models

IEEE TRANSACTIONS ON visualIZATION AND COMPUTER GRAPHICS 2022年第1期28卷 780-790页

作者： Zhao, Zhenge Xu, Panpan Scheidegger, Carlos Ren, Liu Univ Arizona Tucson AZ 85721 USA Amazon AWS AI Seattle WA USA Bosch Res North Amer Cambridge MA USA

The interpretation of deep neural networks (DNNs) has become a key topic as more and more people apply them to solve various problems and making critical decisions. Concept-based explanations have recently become a popular approach for post-hoc interpretation of DNNs. However, identifying human-understandable visual concepts that affect model decisions is a challenging task that is not easily addressed with automatic approaches. We present a novel human-in-the-Ioop approach to generate user-defined concepts for model interpretation and diagnostics. Central to our proposal is the use of active learning, where human knowledge and feedback are combined to train a concept extractor with very little human labeling effort. We integrate this process into an interactive system, ConceptExtract. Through two case studies, we show how our approach helps analyze model behavior and extract human-friendly concepts for different machine learning tasks and datasets and how to use these concepts to understand the predictions, compare model performance and make suggestions for model refinement. Quantitative experiments show that our active learning approach can accurately extract meaningful visual concepts. More importantly, by identifying visual concepts that negatively affect model performance, we develop the corresponding data augmentation strategy that consistently improves model performance.

关键词： visualization data models Analytical models Predictive models Computational modeling Deep learning Task analysis visual data exploration Deep Neural Network Model Interpretation Explainable AI

来源：评论

学校读者我要写书评

暂无评论

ATOVis - A visualisation tool for the detection of financial fraud

引用

INFORMATION visualIZATION 2022年第4期21卷 371-392页

作者： Macas, Catarina Polisciuc, Evgheni Machado, Penousal Univ Coimbra Ctr Informat & Syst Dept Informat Engn Polo 2 Pinhal Marrocos P-3030290 Coimbra Portugal

Fraud detection is related to the suppression of possible financial losses for institutions and their clients. It is a task of high responsibility and, therefore, an important phase of the decision-making chain. Nowadays, experts in charge base their analysis on tabular data, usually presented in spreadsheets and seldom supplemented with simple visualisations. However, this type of inspection is laborious, time-consuming, and may be of little use for the analysis and overview of complex transactional data. To aid in the inspection of fraudulent activities, we develop ATOVis - a visualisation tool that enables a fast analysis and detection of suspicious behaviours. We aim to ease and accelerate fraud detection by providing an overview of specific patterns within the data, and enabling details on demand. ATOVis focuses on applying visualisation techniques to the Finance domain, specifically e-commerce, contributing to the state-of-the-art as the first visualisation tool primarily specialised in Account Takeover (ATO) patterns. In particular, the present paper incorporates: a task abstraction for detecting a specific financial fraud pattern - ATO;two models for the visualisation of ATO;and a multiscale timeline to enable an overview of the data. We also validate our tool through user testing, with experts in fraud detection and experts from other fields of data science. Based on the feedback provided by the analysts, we could conclude that ATOVis is an efficient and effective tool in detecting specific patterns of fraud which can improve the analysts' work.

关键词： Information visualisation finance fraud patterns visual data exploration

来源：评论

学校读者我要写书评

暂无评论

STFT-LDA: An algorithm to facilitate the visual analysis of building seismic responses

引用

INFORMATION visualIZATION 2021年第4期20卷 263-282页

作者： Zhao, Zhenge Motta, Danilo Berger, Matthew Levine, Joshua A. Kuzucu, Ismail B. Fleischman, Robert B. Paiva, Afonso Scheidegger, Carlos Univ Arizona Dept Comp Sci Tucson AZ 85719 USA Univ Sao Paulo Sao Carlos SP Brazil Vanderbilt Univ 221 Kirkland Hall Nashville TN 37235 USA Univ Arizona Dept Civil Engn Tucson AZ 85719 USA

Civil engineers use numerical simulations of a building's responses to seismic forces to understand the nature of building failures, the limitations of building codes, and how to determine the latter to prevent the former. Such simulations generate large ensembles of multivariate, multiattribute time series. Comprehensive understanding of this data requires techniques that support the multivariate nature of the time series and can compare behaviors that are both periodic and non-periodic across multiple time scales and multiple time series themselves. In this paper, we present a novel technique to extract such patterns from time series generated from simulations of seismic responses. The core of our approach is the use of topic modeling, where topics correspond to interpretable and discriminative features of the earthquakes. We transform the raw time series data into a time series of topics, and use this visual summary to compare temporal patterns in earthquakes, query earthquakes via the topics across arbitrary time scales, and enable details on demand by linking the topic visualization with the original earthquake data. We show, through a surrogate task and an expert study, that this technique allows analysts to more easily identify recurring patterns in such time series. By integrating this technique in a prototype system, we show how it enables novel forms of visual interaction.

关键词： visual data exploration time series analysis

来源：评论

学校读者我要写书评

暂无评论

visualization in the preprocessing phase: Getting insights from enterprise professionals

引用

INFORMATION visualIZATION 2020年第4期19卷 273-287页

作者： Paz Milani, Alessandra Maciel Paulovich, Fernando V. Manssour, Isabel Harb Pontificia Univ Catolica Rio Grande Sul PUCRS BR-90619900 Porto Alegre RS Brazil Dalhousie Univ Halifax NS Canada

The current information age has increasingly required organizations to become data-driven. However, analyzing and managing raw data is still a challenging part of the data mining process. Even though we can find interview studies proposing design implications or recommendations for future visualization solutions in the data mining scope, they cover the entire workflow and do not fully focus on the challenges during the preprocessing phase and on how visualization can support it. Moreover, they do not organize a final list of insights consolidating the findings of other related studies. Hence, to better understand the current practice of enterprise professionals in data mining workflows, in particular, during the preprocessing phase, and how visualization supports this process, we conducted semi-structured interviews with 13 data analysts. The discussion about the challenges and opportunities based on the responses of the interviewees resulted in a list of 10 insights. This list was compared with the closest related works, improving the reliability of our findings and providing background, as a consolidated set of requirements, for future visualization research articles applied to visual data exploration in data mining. Furthermore, we provide greater details on the profile of the data analysts, the main challenges they face, and the opportunities that arise while they are engaged in data mining projects in diverse organizational areas.

关键词： visualization preprocessing visual data exploration data mining interviews

来源：评论

学校读者我要写书评

暂无评论

Enabling advanced environmental management via remote and distributed visual data exploration and analysis

引用

COMPUTING AND visualIZATION IN SCIENCE 2012年第3期15卷 123-133页

作者： Krishnan, Hari Meyer, Jorg Romosan, Alex Childs, Hank Bethel, E. Wes Lawrence Berkeley Natl Lab Berkeley CA 94720 USA

We present amethodology that has proven useful in surmounting technical obstacles encountered while producing mobile-enabled, remote visual data exploration and analysis capabilities. Our results focus on a environmental remediation case study. Using a desktop or mobile device, a scientist is able to perform subset selection of a large collection of time-varying observational, well-log data, then to view this data in 3D from arbitrary orientations, and to view output from visualizations of a time-varying 3D fluid flow simulation showing the spread of analytes over time through the same spatial region as the observational data. This infrastructure is capable of delivering imagery to handheld devices at (modestly) interactive rates, supports a variety of different types of mobile devices, and is useful to scientists who may need to access this kind of capability from field locations.

关键词： Remote visualization Distributed visualization visual data exploration visual analysis Mobile devices Environmental management

来源：评论

学校读者我要写书评

暂无评论

A progressive development of a visual analysis interface of climate-related VGI

引用

ENVIRONMENTAL EARTH SCIENCES 2021年第20期80卷 684-684页

作者： Navarra, Carlo Vrotsou, Katerina Opach, Tomasz Joling, Almar Wilk, Julie Neset, Tina-Simone Linkoping Univ Ctr Climate Sci & Policy Res Dept Themat Studies Environm Change Linkoping Sweden Linkoping Univ Dept Sci & Technol Div Media & Informat Technol Norrkoping Sweden Norwegian Univ Sci & Technol NTNU Dept Geog Trondheim Norway Deltares Delft Netherlands

This paper describes the progressive development of three approaches of successively increasing analytic functionality for visually exploring and analysing climate-related volunteered geographic information. The information is collected in the CitizenSensing project within which urban citizens voluntarily submit reports of site-specific extreme weather conditions, their impacts, and recommendations for best-practice adaptation measures. The work has pursued an iterative development process where the limitations of one approach have become the trigger for the subsequent ones. The proposed visual exploration approaches are: an initial map application providing a low-level data overview, a visual analysis prototype comprising three visual dashboards for more in-depth exploration, and a final custom-made visual analysis interface, the CitizenSensing visual Analysis Interface (CS-VAI), which enables the flexible multifaceted exploration of the climate-related geographic information in focus. The approaches developed in this work are assessed with volunteered data collected in two of the CitizenSensing project's campaigns held in the city of Norrkoping, Sweden.

关键词： Geographic visualization Information visualization visual data exploration Volunteered geographic information

来源：评论

学校读者我要写书评

暂无评论

visual exploration of Energy Use in EU 28: Dynamics, Patterns, Policies

引用

ENERGIES 2021年第22期14卷 7532-7532页

作者： Halkos, George Tsilika, Kyriaki Univ Thessaly Sch Econ & Business Dept Econ Lab Operat Res Volos 38333 Greece Hellen Open Univ Sch Social Sci Dept Business Adm Patras 26335 Greece

The paper places emphasis on primary energy resources, their covariation, and their correlation with socioeconomic factors and aims to provide a systematic analysis of their development over time. The analysis uses evidence from European Union (EU) country-level data and is based on visual analytics techniques. Different results from the same territories show that energy consumption does not always reflect or is due to climatological or meteorological conditions. Extensive use of visualization is adopted as a means of contributing to the understanding of energy use, some involved problems and concepts, and energy consumption trends over time. We present an approach that addresses the informatics challenges based on the integration of visualization software, data integration, and cluster analysis. Our cross-sectional energy review advocates that EU energy leaders are moving towards a low-carbon economy. The correlations of energy variables with economic and pollution effects are stronger in greater levels of energy use, which means that energy use has an obvious impact on economic growth and the environment. visual and automated methods employed for the analysis, reveal the direction, the strength, and the nature of the dependence structure, in clusters covering the range of energy use in EU 28 countries.

关键词： energy use visual data exploration correlation analysis trends k-means clustering

来源：评论

学校读者我要写书评

暂无评论

New Analytic Framework of Public Mental Health Prediction Using data Science 1

New Analytic Framework of Public Mental Health Prediction Us...

引用

International Conference on Smart Technologies and Systems for Next Generation Computing (ICSTSN)

作者： Marrapu, Hari Krishna Maram, Balajee Reddi, Prasadu GMR Inst Technol Dept IT Rajam Andhra Pradesh India GMR Inst Technol Dept CSE Rajam Andhra Pradesh India Anil Neerukonda Inst Technol & Sci Dept IT Visakhapatnam Andhra Pradesh India

ISBN: (纸本)9781665421119;9781665421126

Nowadays the human-being is suffering from many issues in the society. In the current situation mental health problem is a prevalence issue. Usage of sensor data and dataset is an interesting issue dealing with prediction of mental health of the people living in society. The data science process can be carried out by collecting datasets from various health sector domains and differentiated into three categories such as low, medium, high. This is helpful in detecting weather the person is suffering from any mental issue or not. A sample test is conducted to assess the health conditions of healthy groups and people having mental illness with the help of data mining in data science field. This paper explains about a novel framework for prediction of public mental health condition which are helpful to handle the challenges in the field of data science. Evaluation of data science surveys is different from traditional data analysis due to the scientific approach method used during data classification and filtration. Here the role of "data Science" framework is for promoting the usage of framework by the healthcare professionals to face challenges of mental health.

关键词： data Science Public Health Mental Health Predictive modeling Pervasive computing visual data exploration data Mining

来源：评论

学校读者我要写书评

暂无评论

A Question-Oriented visualization Recommendation Approach for data exploration 20

A Question-Oriented Visualization Recommendation Approach fo...

引用

Working Conference on Advanced visual Interfaces (AVI)

作者： Lima, Raul de Aranjo Junqueira Barbosa, Simone Diniz Pontificia Univ Catolica Rio de Janeiro Rio de Janeiro RJ Brazil

ISBN: (纸本)9781450375351

The increasingly rapid growth of data production and the consequent need to explore data to obtain answers to the most varied questions have promoted the development of tools to facilitate the manipulation and construction of data visualizations. However, building useful data visualizations is not a trivial task: it may involve a large number of subtle decisions from experienced designers. In this paper, we present an approach that uses a set of heuristics to recommend data visualizations associated with questions, in order to facilitate the understanding of the recommendations and assisting the visual exploration process. Our approach was implemented and evaluated through the VisMaker tool. We carried out two studies comparing VisMaker with Voyager 2 and analyzed some aspects of the recommendation approaches through the participants' feedbacks. As a result, we found some advantages of our approach and gathered comments to help improve the development of visualization recommender tools.

关键词： visualization recommendation visual data exploration visualization tool information visualization

来源：评论

学校读者我要写书评

暂无评论

data Science in Public Mental Health: A New Analytic Framework

Data Science in Public Mental Health: A New Analytic Framewo...

引用

IEEE Symposium on Computers and Communications (ISCC)

作者： Silva, Charith Saraee, Mahsa Saraee, Mo Univ Salford Manchester Sch Comp Sci & Engn Salford Lancs England Univ Manchester Fac Biol Med & Hlth Sch Cardiovasc Sci Manchester Lancs England

ISBN: (纸本)9781728129990

Understanding public mental health issues using data science and finding solutions based on the findings from the data science projects can be complex and requires advanced techniques, compared to conventional data analysis projects. It is important to have a comprehensive project management process to ensure that project associates arc competent and have enough knowledge to implement the data science process. Therefore, this paper presents a new framework that mental health professionals can use to solve challenges they face using data science. Although a large number of research papers have been published on public mental health, few have addressed the use of data science in public mental health. Recently, data Science has changed the way we manage, analyze and leverage data in healthcare industry. data science projects differ from conventional data analysis, primarily because of the scientific approach used during data science projects. One of the motives for introducing a new framework is to motivate healthcare professionals to use "data Science" to address the challenges of mental health. Having a good data analysis framework and clear guidelines for a comprehensive analysis is always a plus point. It also helps to predict the time and resources needed in the early in the process to get a clear idea of the problem to be solved.

关键词： Mental IIealth Public IIealth data Science data Mining visual data exploration

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：