Data Wrangling

Data wrangling generally denotes the cumbersome task of making data useful for analysis. Usually, this means applying hand-crafted scripts, requiring at least a certain degree of technical expertise. In order to make suchlike data preparation accessible for rather casual users as well, we have built a visual analytics prototype in the context of this thesis, easing related tasks. Our approach is an interdisciplinary one, combining contemporary concepts and ideas from various fields: human-computer interaction and usability engineering with information retrieval, data mining, machine learning, plus information visualization. In particular we focus on supporting transformations of time-oriented data since this kind of data exhibits unique characteristics which demand for special consideration. After analyzing related state of the art we identified open issues and derived requirements for such a system. We followed an iterative design process to develop a software prototype called TempMunger, in an agile manner. The process as well as corresponding artifacts are documented and presented in this thesis. Our prototype is a web-based application, tailored for desktop browser usage. More concretely, it offers interactive dashboard visualizations for preferably intuitive and exploratory transformation operations of time-oriented data. A qualitative evaluation of the prototype demonstrates its usefulness and reveals opportunities for future work. Concluding it is safe to say that data wrangling continues to be an exciting field of research where much is yet to be discovered.