MBA

Question 1

Items

Answer 1

objects we are identifying associations between
- group of items : item set

Question 2

Transactions

Answer 2

instances of groups of items CO-OCCURRING together

Question 3

Basket Data

Answer 3

collection of transaction IDs and items bought in a transaction

Question 4

Rules

Answer 4

statements of the form {item1, item2, …} => {itemk}
- implies if an itemset has items on the LHS, visitors will be interested in RHS

Question 5

Support

Answer 5

of an item: fraction of transactions in our data set that contain that item/item set

of a rule: fraction of transactions containing all items in the LHS and RHS item sets (symmetric)

support (A) = #A / #all
support (A->B) = #A&B / #all

Question 6

high support / low support

Answer 6

The higher the support the more frequently the item set occurs.
high support is good
- apply to a large number of transactions; a lot of people affected

Question 7

Confidence

Answer 7

of a rule: probability of transaction that contains the item set on the LHS also contains the item set on the RHS
- proportion of B appearing in transactions that contain A [P(B|A)]

confidence (A->B) = #A&B / #A

Question 8

high confidence / low confidence

Answer 8

The higher the confidence, the greater the likelihood that the item on RHS will be purchased or, in other words, the greater the return rate you can expect for a given rule.

Question 9

Association rule

Answer 9

Statement of the form X->Y = “X implies Y”

Question 10

Lift

Answer 10

ratio by which the confidence of a rule exceeds the expected confidence, assuming that the item sets are independent

lift (A->B) = confidence(A->B) / expected confidence

((expected confidence = support(B) = #B / #all

Question 11

lift value

Answer 11

lift > 1: positive association (items not independent, complementary)
lift = 1: no association (independent)
lift < 1: negative association (items are not independent, substitutes)

Question 12

Rule Notation (using confidence and support)

Answer 12

LHS -> RHS [confidence, support]

Question 13

Association Rule Types

Answer 13

1. Actionable rules: high quality, actionable information
2. Trivial rules: information already well-known by those familiar with the business
3. Inexplicable rules: no explanation and do not suggest action

*trivial and inexplicable occur the most often

Question 14

Threshold

Answer 14

1. Support threshold
   - eg. set threshold for minimum support => all item sets equal and above are frequent item sets
2. Confidence threshold
   - eg. set minimum confidence for an identified rule

Question 15

Apriori principle

Answer 15

if item set is frequent, all of its subsets must also be frequent

Question 16

Apriori Algorithm

Answer 16

• starts from single items
• progressively identifies larger items subsets of different sizes
• exploits Apriori Principle that any subset of a frequent item set is also a frequent item set
• any superset of an infrequent item is also infrequent

example: Algorithm to generate frequent item sets
1. Start with all item sets with a single item and compute their support (or count)
2. Remove the ones that do not have minimum support
3. Generate all two-item item sets using the results retained in the previous step and compute their support
4. Remove the ones that do not have minimum support
5. Continue till you have item sets of all sizes (or desired max size) with the required minimum support (or count)

Question 17

Apriori Parameters

Answer 17

Examples:
- maximum number of items to process
- support threshold
- confidence threshold
- maximum size of item set
- maximum number of rules to keep