TPDA2 ALGORITHM FOR LEARNING BN STRUCTURE FROM MISSING VALUE AND OUTLIERS IN DATA MINING
ABSTRACT: Three-Phase
Dependency Analysis (TPDA) algorithm was proved as most efficient algorithm
(which requires at most O(N4) Conditional Independence (CI) tests). By
integrating TPDA with “node topological sort algorithm”, it can be used to
learn Bayesian Network (BN) structure from missing value (named as TPDA1
algorithm). And then, outlier can be
reduced by applying an “outlier detection & removal algorithm” as
pre-processing for TPDA1. TPDA2 algorithm proposed consists of those ideas,
outlier detection & removal, TPDA, and node topological sort node.
Author: Benhard Sitohang, G.A.
Putri Saptawati
Journal Code: jptinformatikagg060005
