Seeing association rules
There are many situations where we're interested in patterns demonstrating the co-occurrence of some items. For example, marketers want to know which goods are often bought together, clinical personnel need to know symptoms associated with certain medical conditions, and in information security we want to know which activity patterns are associated with intrusion or fraud. All of these problems have a common structure: there are items (goods, symptoms, records in logs) organized in transactions (shopping list, medical case, user activity transaction). With this type of data, we can then analyze it to find association rules, such as If the client bought a lemon and some cookies, he is also likely to buy tea, or in more formal notation: (cookies, lemon → tea).
Note
We will use pictograms throughout this chapter to facilitate the visual notation of item sets and rules: {
→
}.
These rules allow us to make informed decisions, such as putting associated items on the same...
