Making scientific discoveries based on large and heterogeneous datasets is challenging. The continuous improvement of data acquisition technologies makes it possible to collect more and more data. However, not only the amount of data is growing at a fast pace, but also its complexity. Visually analyzing such large, interconnected data collections requires a user to perform a combination of selection, exploration, and presentation tasks. In each of these tasks a user needs guidance in terms of (1) what data subsets are to be investigated from the data collection, (2) how to effectively and efficiently explore selected data subsets, and (3) how to effectively reproduce findings and tell the story of their discovery.
On the basis of a unified model called the SPARE model, this thesis makes contributions to all three guidance tasks a user encounters during a visual analysis session: The LineUp multi-attribute ranking technique was developed to order and prioritize item collections. It is an essential building block of the proposed guidance process that has the goal of better supporting users in data selection tasks by scoring and ranking data subsets based on user-defined queries. Domino is a generic visualization technique for relating and exploring data subsets, supporting users in the exploration of interconnected data collections. Phoeva is a novel open-source visual analytics platform designed to speed up the creation of domain-specific exploration tools. The final building block of this thesis is CLUE, a universally applicable framework for capturing, labeling, understanding, and explaining visually driven exploration. Based on provenance data captured during the exploration process, users can author "Vistories", visual stories based on the history of the exploration. The practical applicability of the guidance model and visualization techniques developed is demonstrated by means of usage scenarios and use cases based on real-world data from the biomedical domain.