COMBINATION OF MULTIPLE CLASSIFIERS USING BAGGING IN SEMI-SUPERVISED LEARNING

摘要

PURPOSE: a kind of device for data mining model total figure because providing a kind of classifier for improving precision of prediction by ensemble learning by using label and unlabeled data prediction label using unlabeled data. ;CONSTITUTION: it is prediction semi-supervised learning method (S601) that unlabeled data, which is selected from unlabeled data and label around unlabeled data,. The labeled data that basic classification device carries out unlabeled data is divided into training set (S602). The basis that overall model is established is by assembled classifier (S603). The prediction of the label unlabeled data provides the data based on confidence level, by converting LNP (linear neighborhood propagation). ;The 2013 of copyright KIPO submissions;[Reference numerals] (AA) starts; (BB) terminate; (S601) stage of the selection data from a unlabeled data group prediction label is selected; (S602) prediction label is formulated to be divided into training set and labeled data with corresponding unlabeled data and generate supervised study classifier; (S603) combined classification model is generated by repeating S601 and S602 in conjunction with what supervised study classifier obtained