Mining MOUCLAS Patterns and Jumping MOUCLAS Patterns to Construct Classifiers - Data Mining: Theory, Methodology, Techniques, and Applications - page 126

The first step of the MOUCLAS-2 algorithm is given in Figure 3 as follows.
1 X = genJEP (I); // generate all the JEPs of all the class y in D
2 for each class y do
3 for each JEP of a same class y do
4 Xi = reduceDim (I); // reduce the dimensionality on the set of all items I in D
labeled with class y based on the attributes of the JEP
5 E i = genClusterrules( cluset, class ); // generate a set of cluster_rules, namely
JMPset, based on X i
6 for each transaction d ∈ C do

E i can be supported by d then e.cluCount++ ; // accumulate the
cluCount of cluster_rule e supported by d
8 end
9 subsup i =

7 if one e ∈

e.cluCount ; //calculate the subsup of each JMPset

C

10 end
11 end
12 JMPs = ∪ E i ; // discover the final set of JMP

Fig. 3. The Training Phase of the MOUCLAS-2 Algorithm

1 for each transaction d ∈ D do

2 for each class y do
3 for each JMPset of a same class y do
4 if d satisfies a JMPset then e.subsupt++ ; // accumulate the subsup of JMPsets
supported by d
5 end
6 the subsup y of d in class y = e.subsupt ; // calculate the total subsup of d in
class y
7 end
8 if subsup y is the maximum then d is labeled as y
9 if the subsup in two or more classes are the same then d is labeled as the class,
whose JMPs are generated earlier than the others .
10 if the subsup = 0 then d is labeled as a default class
11 end

Fig. 4. The Testing Phase of the MOUCLAS Algorithm

In the testing phase, The MOUCLAS-2 algorithm also consists of two sub-steps, by
which the J-MP classifier can classify test data:

Algorithm: Classification Process of J-MP Classifier
Input: A test database, D ; The set of Jumping MOUCLAS patterns ( JMPs ); The
support of transaction d belong to JMPs in C ( subsup )
Output: classification result of test database
Methods:
(1) Determine the subsup of each transaction d in D in each class.
(2) Classify the test data.