Week 10 Flashcards
What is Association Rule Mining (ARM)?
A “rule”-based machine learning method
What is the objective of ARM?
To discover interesting relations between variables in large DBs.
Association Rule Mining aims to identify _____ rules discovered in DBs, using some measures of interestingness
strong
T/F - Association Rule Mining assumes all data is numeric
False. ARM assumes all data is categorical
T/F - ARM features good algorithms for numeric data
False
What was Associated Rule Mining initially used for?
Market analysis, to determine how items purchased by customers are related
A customer buys bread, there is an 80% chance they buy butter.
bread > butter
or
A customer buys {onions, potatoes} > buys{tomatoes}
What is this an example of?
Associated Rule Mining
It was useful for marketing activities like product promotion or product pricing
Bread > Butter[20%, 45%]
What is the Antecedent, Consequent?
What is the Support, Confidence?
Antecedent = Bread
Consequent = Butter
Support = 20%
Confidence = 45%
In a store where there are 100 total transactions, what does a support of 20% for bread mean?
It means 20 transactions contained bread
With 20 transactions that contain bread there is a 45% confidence that the transaction also contained butter. How many transactions had butter?
9
T/F - Recommendation systems like Amazon’s “also viewed” use ARM
True
Here are 3 algorithms for association rule:
Apriori Algorithm
Elcat Algorithm
F.P. Growth Algorithm
Which one is the best?
Apriori
What do we find with this equation?
of transactions where X appears / Total # of transactions
Support (X)
To find this we divide Support(X,Y) with Support (X)
Confidence
Algorithm for k products:
- User sets a minimum support criterion
- Generate a list of one-item sets that meet the support criterion
- Use the list of one-item sets to generate list of two-item sets
- Use the list of two-item sets to generate list of three-item sets
- Continue up through k-item set
- All item sets must meet the support criterion
What do we get by following this algo.?
The Apriori algo. for Generating Frequent Item sets
What is the maximum Confidence value for a X > Y rule if the consequent and antecedent always occur together? What is the minimum?
1, 0
What does Lift measure?
How often X and Y happen together vs how often they happen independently
What does it mean if the lift is greater than 1?
There is a positive correlation
What does it mean if the lift is less than 1?
There is a negative correlation
What does it mean if the lift is 1?
There is no relationship
Association rules produce rules based on associations between ______ from a ___________ of transactions
items from a database of transactions
