A methodology for the visual comprehension of Big Data.
Big Data Analytics is continuously growing, especially since the last decade, and Visual Analytics have become a key component in order to analyze it. However, defining analytical goals and using the most suitable visualizations is a complex task, especially for non-expert users in data visualization. Consequently, it is possible that graphics are misinterpreted, contributing to making wrong decisions that lead to missed opportunities.
Therefore, the main goal of this doctoral thesis is to define a methodology that groups a series of techniques and approaches to improve the visual understanding of Big Data. Specifically, the current needs in the taking of requirements for the generation of visualizations have been analyzed and a complete methodology has been proposed, from the definition of the user requirements to the implementation of the visualizations. This methodology guides the user in the definition of their analytical goals and automatically generates the best suited visualization for each goal by grouping them into dashboards.
The methodology is composed by; (i) a User Requirements Model, (ii) a Data Profiling Model that semi-automatically extracts information about the characteristics of the data sources, and (iii) a Data Visualization Model. Our proposal has been evaluated and applied in different areas such as: Smart Cities, industrial production processes, and health environments. Furthermore, with the results obtained and presented in this work, we can conclude that the main goal of this doctoral thesis is achieved since, in line with the experiments carried out in the core of this doctoral thesis, our proposal; (i) allows users to cover more analytical questions, (ii) improves the set of generated visualizations, and improves (iii) the overall satisfaction of the users.
As a result of the research carried out in this doctoral thesis, numerous scientific articles have been obtained and presented in international congresses and high impact journals. That is why this doctoral thesis is presented by a compendium of articles.