文献类型：专利

英文题名：Category equalization based training data sample discovering method, involves selecting training data sample by user, and determining training data sample by performing sample discovery operation according to corresponding category subset

作者：YUAN Y;GU Y;TAN X;RUAN T

机构：[1]UNIV EAST CHINA SCI & TECHNOLOGY

申请号：CN108062563-A

申请日：2017-12-12

公开日：2018-05-22

语种：英文

收录：DERWENT

摘要：NOVELTY - The method involves pre-processing candidate set input in an original database, where the input comprises missing value and abnormal data value. Training data sample is selected by a user based on category equalization. The training data sample is determined by performing standard sample discovery operation according to a corresponding category subset. The missing value is deleted from the original database. The abnormal data value is identified and deleted according to attribute of pattern data. The corresponding category subset is divided according to categories. USE - Category equalization based training data sample discovering method. ADVANTAGE - The method enables improving classification accuracy and machine learning efficiency of an intellectual classifying system. DESCRIPTION OF DRAWING(S) - The drawing shows a flowchart illustrating a category equalization based training data sample discovering method. '(Drawing includes non-English language text)'