PENGEMBANGAN LEARNING CHARACTERISTIC RULE PADA ALGORITMA DATA MINING ATTRIBUTE ORIENTED INDUCTION

Abstract

This paper shows the improvement of current characteristic rule learning in Attribute Oriented Induction (AOI) data mining technique. The proposed algorithm was applied with improvement upon current algorithm with 3 steps where the first step is elimination for checking condition if there is no higher level concept in concept hierarchy for attribute. The second step is elimination of attribute removal if fulfill for checking condition if there is no higher level concept. The third step is elimination of attributes in input dataset which no higher level concept in concept hierarchy. The development of these data mining algorithm applied Knowledge Data Discovery (KDD) methodology which consist 7 steps. Current and proposed AOI characteristic rule learning were implemented with server programming such as PHP Hypertext Preprocessor (PHP) and using 4 input datasets such as adult, breast cancer, census and IPUMS from University of California, Irvine (UCI) machine learning repository. The experiments showed that proposed AOI characteristic rule are better than current AOI characteristic rule, where experiments upon adult, breast cancer, census, IPUMS datasets have average 11, 3.8, 7.2, 7.2 respectively times better performance. The experiments were carried on AMD A10-7300(1.90 GHz) processor with 8.00 GB RAM