Exploring New Ways of Utilizing Automated Clustering and Machine Learning Techniques in Information Visualization

Abstract

Research Area: Information visualization, human-computer interaction.Research Topic. The main research topic of the thesis is to explore the possibilities of automated clustering and machine learning techniques for developing new approaches in information visualization.Research Problem. The main goal of information visualization is to present data to the users in a way that optimizes intelligibility of the data and support the detection of relevant patterns in the data, where the application context defines what qualifies as 'relevant'. Many different approaches typically tailored to a specific problem have been developed within the past years. At the same time the application of mathematical methods for data analysis and identification of patterns has substantially increased, and is typically referred to as data mining. Different visualization techniques are used in data mining, however the systematic and dynamic integration of data mining techniques with visualization approaches is only in its beginning.The main research problem tackled in the thesis is how to best integrate approaches from traditional data mining and information visualization domains to achieve more usable and helpful analysis tools for large amounts of data. The basic visualization approach explored is tag clouds, and especially concepts in displaying them according to a semantic structure as proposed e.g. by [2,3,1] are further developed. The overall research problem can be divided into sub-problems which need to be addressed in order to develop actually useful approaches:• Develop new approaches and concepts for integrating data mining and information visualization techniques • Develop a detailed understanding of the human perception and analysis process involved when interacting with these systems • Develop a understanding on the impact of errors, artifacts and imperfection in the data analysis on experienced quality and usefulness • Define quality benchmarks that are needed to be met to provide real value and benefit to the user when using such systems • Evaluate the developed concepts and prototypes in realistic scenarios, to be able to estimate their value and applicability in real-world contexts.