ANALYZING METHOD AND REPRESENTING METHOD FOR DATA AREA
申请公布号:JP2001250101(A)
申请号:JP20000061709
申请日期:2000.03.07
申请公布日期:2001.09.14
发明人:KASEDA OSAO;TSUTSUI HIROAKI
分类号:G06N3/00;G06T7/00;G10L15/06;G10L15/16;(IPC1-7):G06N3/00
主分类号:G06N3/00
摘要:PROBLEM TO BE SOLVED: To make it easy to grasp how much data groups belonging to the same class vary, select the best feature space, and generate the best categorization model. SOLUTION: When (n)-dimensional data which belong to one class in an (n)-dimensional feature space defined by (n) kinds of variables and have their position specified with the variables are inputted, the feature space is divided into mn areas by dividing the respective variable by (m) and a divided area including the (n)-dimensional data is decided as a learning area belonging to the class through a learning area generating process (step 101); and learning areas in one data area as which a group of connected learning areas is regarded are given the same label through a labeling process (step 102) and proximate areas are set and displayed by the data areas given the same label through an approximating process (steps 103 and 104).