Data Mining for Business Analytics

CHAPTER 3 Data Visualization

In this chapter, we describe a set of plots that can be used to explore the multidimensional nature of a data set. We present basic plots (bar charts, line graphs, and scatter plots), distribution plots (boxplots and histograms), and different enhancements that expand the capabilities of these plots to visualize more information. We focus on how the different visualizations and operations can support data mining tasks, from supervised tasks (prediction, classification, and time series forecasting) to unsupervised tasks, and provide a few guidelines on specific visualizations to use with each data mining task. We also describe the advantages of interactive visualization over static plots. The chapter concludes with a presentation of specialized plots suitable for data with special structure (hierarchical, network, and geographical).

3.1 Introduction¹

The popular saying “a picture is worth a thousand words” refers to the ability to condense diffused verbal information into a compact and quickly understood graphical image. In the case of numbers, data visualization and numerical summarization provide us with both a powerful tool to explore data and an effective way to present results (Few, 2012).

Where do visualization techniques fit into the data mining process, as described so far? They are primarily used in the preprocessing portion of the data mining process. Visualization supports data cleaning by finding incorrect values (e.g., patients whose ...