Data-mining HW

Association Rules (FP-Growth)

Tid Items
100 A, C, D
200 B, C, E
300 A, B, C, E
400 B, E

Strong rule
{B, E}->C (2/3)
C->A (2/3)
A->C (2/2)

Sequential Pattern (PrefixSpan)

SID Sequences
100 <(1,5) (2) (3) (4)>
200 <(1) (3) (4) (3,5)>
300 <(1) (2) (3) (4)>
400 <(1) (3) (5)>
500 <(4) (5)>

Some frequent sequential patterns
{1,2,3,4}, {1,3,5}, {4,5}

Classification (decision tree with gini index)

@attribute marital_status {S,M}
@attribute num_children_at_home numeric
@attribute member_card {Basic,Normal,Silver,Gold}
@attribute age numeric
@attribute year_income numeric

Example database:
  {0 M,1 1,2 Silver,3 82,4 40000}
  {0 M,1 2,2 Silver,3 53,4 40000}
  
{0 M,1 1,3 59,4 60000} member_card = basic
{1 3,2 Silver,3 43,4 160000} member_card = basic
{0 M,1 2,3 52,4 120000} member_card = silver 

Clustering(Dbscan)

Input: given a node with two dimensional attributes X and Y 
Output: group these nodes into several clusters. The format is X Y ClusterID 
Example 
Input 2 3 (X=2, Y=3)
Output 2 3 1(X=2 Y=3 ClusterID=1)

There are five test data in the attachments
Each row is a data point with x and y value
Hint: 
Test1-3 should be clustered to 4 clusters