1998 Fiscal Year Final Research Report Summary
Logical Analysis of Data and Knowledge Acquisition
| Project/Area Number |
09044160
|
|---|
| Research Category |
Grant-in-Aid for international Scientific Research
|
| Allocation Type | Single-year Grants |
|---|
| Section | Joint Research |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | KYOTO UNIVESITY |
|---|
Principal Investigator |
IBARAKI Toshihide Kyoto Univ., Graduate School of Informatics, Professor, 情報学研究科, 教授 (50026192)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
HAMMER Peter Rutgers University, RUTCOR,Professor, ラトコー研究所, 教授
BOROS Eudre Rutgers University, RUTCOR,Professor, ラトコー研究所, 教授
MASUYAMA Shigeru Toyohashi Inst.of Tech.Science, Faculty of Eng, Prof., 工学部, 教授 (60173762)
YAGIURA Mutsunori Kyoto Univ., Graduate School of Informatics, Assis.Prof., 情報学研究科, 助手 (10263120)
NAGAMOCHI Hiroshi Graduate School of Informations, Associate Prof., 情報学研究科, 助教授 (70202231)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | Logical Analysis of Data / Knowledge Aquisition / Data Mining / Knowldge Discovery / Booleam functions |
|---|
| Research Abstract |
The main purpose of this research is to extract meaningful information from the data of large size, which can be found in various fields of this modern society. In particular, we are interested in the method which is called the logical analysis of data (LAD). To acquire the knowledge from the data set, LAD constructs a Boolean function consitstent with the given data, and then analyzes it from the logical point of view.
Mathmatically, a data set consists of a set of positive examples T * {O, i} ^n and a set of negative examples F * {O, 1}^n . A Boolean function f is called an extension of (T, F) if f outputs 1 (resp., 0) for every vector in T (resp., F). In this study, we clarified the conditions on (T, F) so that its extension exists in a given class of functions C.As data usually contain errors and/or incomplete bits, the concept of extensions is then generalized to cope with such situations, and the conditions for the exsistence of generalized extentions are throughly investigated.
Also, many real world data contain numerical attributes. These data can be transformed to binary data, by introducing appropriate cut points. The complexity and algorithms for the problem of minimizing the number of necessary cut points are studied and settled in this research.
|
