Visualization and analysis of probability distributions of large temporal data

Increasingly, data is recorded at much finer temporal scales. However, data collected at an hourly scale can also be analyzed using coarser scales such as days, months or quarters. Cyclic granularities representing repetitions in time (such as hour-of-the-day, day-of-the-week, work-day/weekend) are effective for analyzing repetitive patterns in time series data. To fully comprehend these patterns, one must traverse all cyclic granularities. This is difficult with many options but only few of them revealing major patterns. This thesis presents methods for screening the interesting ones and then visualizing the distributions to support the discovery of regular patterns and clusters of behaviours.