2. What is Association Mining?
Able to narrow down “rules” to specific products to highlight rela-onships
Products that are used together are grouped together by transac-on types (Users)
Rela-onships are then discovered between product usage
3. Metrics Used
Support: The percentage of transac-ons that contain all of the items in an itemset
Larger the support, the more sta-s-cally sound
Confidence: The probability that a transac-on that contains the items on the leF side of the rule also contains the item
on the right hand side
Larger the confidence, the greater likelihood the the item on the right side will be played
Li7: The probability of all of the items in a rule occurring together divided by the product of the probabili-es of the items
on the leF and right hand side occurring as if there was no associa-on between them
Larger the liF, the greater the link between the two products
Source: h*ps://select-sta2s2cs.co.uk/
5. How to Read Rules
Users who play Scratch and Win (Android) are likely to play….
* All Android related platforms. **All greater than 12% support
Confidence
Lift
46%
180%
Ø If someone plays SNW, they are 46% likely to play Unlock & Win
Ø If someone plays SNW, their chance of playing Unlock & Win would
increase by 80%
6. Top Rules
Users who play Scratch and Win (Android) are likely to play….
* All Android related platforms. **All greater than 12% support
Confidence
Lift
46% 46% 40% 31% 30%
180% 114% 138% 137% 147%