Thesis (M.Sc.) - Asian Institute of Technology, 2011

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science, School of Engineering and Technology

Exploratory Data Analysis explored data set by using visualization and data manipulation in order to reveal the underlying patterns and relationship among data. Today{u2019}s interfaces do not typically support the capture or review of the states in this process, which could answer how the analysts came to their conclusions. Additionally, in order to collaborate amo ng analysts, they often require a collection set of exploration states to be shared as well as the relevant domain ata. As a result, the data analysis process itself is necessary for recording in order to be able to freely navigate or backtrack to the analysis activities, and then perform further analysis based on an interesting point in time or explore alternative exp loration paths.The idea to move back and forth between states can be found in researches since long ago. However, most of these researches targets specific application domain, and yet few generalized architecture supporting those idea has ever been p roposed. As a result, few of the application benefit from those researches. The purpose of this thesis is to propose a generalized architecture used for bookmarking and replaying states across an analysis activity.The primary attribute of the architect ure would promote high flexibility that allows many application domains to actually adopt and apply it. Furthermore, the performance of different ways of bookmarking and replaying states, which includes state copying and command pattern , were also compare d and evaluated toward s determining the most suitable architecture. Our evaluation results show that our architecture promotes high flexibility and could be deployed in many application domains. A future expansion for this architecture c ould support multi ple timeline, which would enable parallel exploration of the solution space .