Association Analysis

Preprocessing

I clean the data by dropping data with no InvoiceNo or InvoiceNo begin with C or not located in United Kingdom or the item is POSTAGE

Implementation

pseudocode

for each case X ⇒ Y:
	support = P(X & Y)
	confidence = P(X & Y) / P(X)
	if support > 0.01 and confidence > 0.5:
		they are associated

If confidence > 0.5, that means for all people who buy X, the number of people who buy Y is more than the number of people who do not buy Y. But if confidence > 0.5 and support.