Data Mining concepts and technologies chapter 2 (6th) Mining of Large Databases

Project set: the set of items, which is recorded as I. Milk, bread, apple, etc transaction: Transaction T is a set of projects in I, each transaction has a TID as the identifier. Project set X support Sup (X) Count (X) | D | Associate Rule X-Y support Sup (X-Y) Count (X-Y) | D | Associate Rule X-Y confidence level Conf

Project set: the set of items, which is recorded as I. Milk, bread, apple, etc: Transaction T is a set of projects in I. Each transaction has a TID as the identifier. Project set X support Sup (X) = Count (X)/| D | Associate Rule X-Y support Sup (X-Y) = Count (X-Y) /| D | confidence level Conf of the association rule X-Y

Project set: A set of items, which is counted as I. Milk, bread, apple, etc

Things ServicesTransaction T is a collection of projects in I. Each transaction has a TID as the identifier.

Project setXSupported

Sup (X) = Count (X)/| D |

Association RulesX-> YSupported

Sup (X-> Y) = Count (X-> Y)/| D |

Association RulesX-> YConfidence Level

Conf (X-> Y) = Count (X-> Y)/Count (X)

Steps:

1. Find all frequent item sets

2. Frequent Item sets generate strong Association Rules

AlgorithmAprioriAlgorithm

Core Ideology

Frequent 1-then association rule then pruning
Then frequency 2-then association rule then pruning
Then frequency 3-then association rule then pruning
......

Repeat this process until you cannot proceed.

Algorithm ImplementationWrite it separately later

Improvement MeasuresFP (Frequent Pattern Growth) Frequent mode Growth

Multi-layer Association Rules

Basic Idea:

Search for frequent item sets at each conceptual layer from top to bottom. Then ① explore association rules at this layer ② appropriate pruning, and go down to a more specific layer.

Optional solution:

1. Use consistent minimum support

2. Use the minimum support degree of decline

3 independent layer by layer

Layer-4 cross-item Filtering

5. Cross k-item set filtering

Multidimensional Association Rules

Association rules involving two or more dimensions

E.g. age (X, "IBM desktop computer") ^ occupation ("drive" r) => buys (X, "laptop ")

Attention to Multidimensional Association Rules

Strong association rules are not necessarily interesting. For example,

Buys (X, "computer games") => buys (X, "videos") [Support = 40%, confidence = 66%]

The derived results may meet the support and confidence level, but they are not interesting.

In fact, there may be 75% video users who do not purchase computer games, but only 66% video users who have bought computer games and videos at the same time, which indicates that computer games and videos are negatively correlated.