Social Media Mining: An Introduction

P1: Sqe Trim: 6.125in×9.25in Top: 0.5in Gutter: 0.75in
CUUS2079-05 CUUS2079-Zafarani 978 1 107 01885 3 January 13, 2014 19:23

106 Data Mining Essentials

Data

Selection

Preprocessing

Preprocessed

Data

100

50

18 94

3

76

1

94

3

76

1

Low

Med

Hi

2 Low

Transformed

Data Patterns

Target Data

Transformation

Interpretation or

Evaluation

Knowledge

Data Mining

Figure 5.1. Knowledge Discovery in Databases (KDD) process.

To analyze social media, we can either collect this raw data or use

available repositories that host collected data from social media sites.^1

When collecting data, we can either use APIs provided by social media sites

for data collection or scrape the information from those sites. In either case,

these sites are often networks of individuals where one can perform graph

traversal algorithms to collect information from them. In other words, we

can start collecting information from a subset of nodes on a social network,

subsequently collect information from their neighbors, and so on. The data

collected this way needs to be represented in a unified format for analysis.

For instance, consider a set of tweets in which we are looking for common

patterns. To find patterns in these tweets, they need to be first represented

using a consistent data format. In the next section, we discuss data, its

representation, and its types.

5.1 Data

In the KDD process, data is represented in atabularformat. Consider the

example of predicting whether an individual who visits an online book

seller is going to buy a specific book. This prediction can be performed

by analyzing the individual’s interests and previous purchase history. For

instance, when John has spent a lot of money on the site, has bought similar

books, and visits the site frequently, it is likely for John to buy that specific

book. John is an example of aninstance. Instances are also calledpoints,

data points,orobservations.Adatasetconsists of one or more instances: