Association Rules -- July 10, 2022

Association rules

https://zhuanlan.zhihu.com/p/118439868

Apriori Algorithm — Classical algorithms

Look what the person who bought this product bought with others .

Relevant concepts

Transaction set ： Put all the goods together .
Record ： A single number means a record
project （ term ）： A, B,C A single item
Project set （ Itemsets ）： A collection of items
K Itemsets ： The number of elements in the itemset is K,

The following is the content of probability theory ：

Def 6 Support （support ）

Sup(x) = （ The number of times an itemset occurs in a transaction ）/ The total number of records in the transaction set
It can be understood as probability （ frequency ）.

Def 7 Degree of confidence （ Conditional probability ）

stay x On the basis of ,Y Probability of occurrence .

Definition ：
Promotion ： A Is the emergence of right B The emergence of has improved .
lift (A->B) If it is greater than 1, shows B stay A Send based on B It's more likely to happen alone .
be A Appear right B Occurrence has a positive effect .

Two theorems

Theorem 1: If X It's a frequent K Itemsets , Then all its subsets must be frequent
Theorem 2： If X No k-1 Item frequency , Then it must not be frequent K Itemsets

Algorithm flow

Here is an example ：

from K-1 To K term , Combinations cannot be randomly grouped , Want to have K-2 The items are the same
Pre pruning added

There are related relationships below
Analyze algorithm performance