Design And Implementation Of An Data Visualization And Exploration System

In the big data era, the ability for analyzing massive dada is becoming more and more important. Discovering valuable information from massive data largely depends on the experience and analyzing skills of human, while a data visualization and exploration system can be helpful to improve users’analyzing ability by generating carefully designed visual representation of the data. With the help of these systems, user can better verify their conjectures about the data set and dig out the underlying valuable information. On the other hand, with the increasing size of the data set and the number of visual primitives, these systems are facing scalability challenges.In this dissertation, we explain the generation process of important types of visual diagrams in detail, and analyze the key point of different diagrams. Among others, we present the implementation details of layout algorithm, pseudo-code of tree diagram and word cloud diagram, and introduce the exponential smoothing method for forecasting time-series data and present a generalized way of calculating different smoothing models. In the last part of this dissertation, we discuss two optimizations for the diagram generating process and diagram data transferring process, and present the evaluation of these optimizations.Experiments show that our system provides better performance and scalability than the existing open-source visualization toolkits, such as d3.js, thus can be used to efficiently explore larger dataset. We also present a case study to evaluate the usability of the system.