Clustering Methodology for Symbolic Data

2Symbolic Data: Basics

In this chapter, we describe what symbolic data are, how they may arise, and their different formulations. Some data are naturally symbolic in format, while others arise as a result of aggregating much larger data sets according to some scientific question(s) that generated the data sets in the first place. Thus, section 2.2.1 describes non‐modal multi‐valued or lists of categorical data, with modal multi‐valued data in section 2.2.2; lists or multi‐valued data can also be called simply categorical data. Section 2.2.3 considers interval‐valued data, with modal interval data more commonly known as histogram‐valued data in section 2.2.4. We begin, in section 2.1, by considering the distinctions and similarities between individuals, classes, and observations. How the data arise, such as by aggregation, is discussed in section 2.3. Basic descriptive statistics are presented in section 2.4. Except when necessary for clarification purposes, we will write “interval‐valued data” as “interval data” for simplicity; likewise, for the other types of symbolic data.

It is important to remember that symbolic data, like classical data, are just different manifestations of sub‐spaces of the ‐dimensional space always dealing with the same random variables. A classical datum ...