Using Association Rules for Classification from Databases Having Class Label Ambiguities: A Belief Theoretic Method - Data Mining: Foundations and Practice - page 534

Tabl e 3 . The generated rule set

Outlook Humidity Windy Decision Support Confidence

LOW Play 0.5556 1.0000

sunny MEDIUM Play 0.4444 1.0000

sunny LOW Play 0.3333 1.0000

sunny FALSE Play 0.3333 1.0000

LOW TRUE Play 0.3333 1.0000

sunny Play 0.7778 0.8750

sunny TRUE Play 0.4444 0.8000

MEDIUM Play 0.4444 0.6667

TRUE Play 0.5556 0.5556

FALSE Play 0.4444 0.5000

rainy FALSE Don't play 0.4000 1.0000

HIGH FALSE Don't play 0.4000 1.0000

rainy Don't play 0.4000 0.6667

overcast HIGH Don't play 0.4000 0.6667

overcast FALSE Don't play 0.4000 0.6667

FALSE Don't play 0.8000 0.5000

overcast Don't play 0.6000 0.5000

HIGH Don't play 0.6000 0.5000

MEDIUM FALSE Don't play 0.4000 0.5000

MEDIUM Don't play 0.4000 0.3333

0.3333 1.0000

sunny HIGH Θ C

rainy TRUE Θ C 0.3333 1.0000

rainy HIGH TRUE Θ C 0.3333 1.0000

sunny HIGH TRUE Θ C 0.3333 1.0000

HIGH TRUE Θ C 1.0000 0.7500

HIGH Θ C 1.0000 0.5000

rainy HIGH Θ C 0.3333 0.5000

overcast HIGH TRUE Θ C 0.3333 0.5000

TRUE Θ C 1.0000 0.3333

rainy Θ C 0.3333 0.3333

overcast HIGH Θ C 0.3333 0.3333

overcast TRUE Θ C 0.3333 0.3333

sunny TRUE Θ C 0.3333 0.2000

overcast Θ C 0.3333 0.1667

sunny Θ C 0.3333 0.1250

3 Experimental Results

Several experiments on two different groups of databases (databases with
and without class label ambiguities) were conducted to compare the perfor-
mance of our proposed ARM-KNN-BF classifier with some existing classifica-
tion methods.