Events Analysis in Visual Analytics

Abstract

Data production is in constant exponential growth in various domains. The understanding of big data becomes a central competitive aspect. Recently, ownership and usage of data have been discussed on an international scale considering its power and possible consequences for society. Its full potential impact remains unknown. Data awareness can benefit people from the personal to the governmental level. Successful data analysis support decision-making in different scaling and life aspects. Regardless of the domain and scale, most data is being collected and treated present multivariate and time-oriented aspects.Event analysis takes into consideration variables that change behavior over time. Data pattern and data anomalous identification and the reasoning about it support critical tasks among various domains. Currently, event analysis solutions use mainly data mining approaches. However, applying Visual Analytics (VA) techniques may enhance the knowledge discovery process and increase the detection and prediction of events’ accuracy. As displaying distinct data perspectives in multiple views and with interactive support, VA aspects allow users to get familiar with the data while exploring it. By coupling human visual perception skills and domain knowledge, VA presents improved cognitive advantages.We propose to investigate how VA can be applied to tackle the main challenges in event analysis. The main contributions of this thesis are: (1) we developed distinct VA approaches in close collaboration with experts from different domains to support real-world datasets and improve event analysis tasks from their existing workflow, (2) we present the first VA approach based on a scoring system for financial fraud events detection, (3) we offer a new guidance-enriched component for network pattern generation, detection, and filtering that supports different levels of analysis complexity, (4) we conducted different evaluations of our solutions that presented positive results, and (5) we elaborate on possible future research directions and open challenges in the field. All of our discoveries have been collected through continuous collaboration with different domain experts during each experiments’ design, development, and evaluation.