2004 Fiscal Year Final Research Report Summary
Data gathering and preprocessing apprentice
| Project/Area Number |
13480090
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Osaka University (2003-2004)
Tokyo Institute of Technology (2001-2002) |
|---|
Principal Investigator |
NUMAO Masayuki Osaka University, Institute of Scientific and Industrial Research, Professor, 産業科学研究所, 教授 (30198551)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
NATTEE Cholwich Osaka University, Institute of Scientific and Industrial Research, Research Associate, 産業科学研究所, 助手 (30379101)
SATO Taisuke Tokyo Institute of Technology, Graduate School of Information Science and Engineering, Professor, 大学院・情報理工学研究科, 教授 (90272690)
|
|---|
| Project Period (FY) |
2001 – 2004
|
| Keywords | machine learning / agent / artificial intelligence / data mining |
|---|
| Research Abstract |
This project proposed and implemented an automatic information gathering and preprocessing system for data mining. Information gathering and preprocessing are usually a cooperative work by an information source, domain experts and mining experts. They often have to exchange and update their data. If we use an email, we will soon have a full of emails. To use a Web page we have to design an appropriate page. To overcome these difficulty, the investigators propose a new tool, WAVE (Word-of-mouth-Assisting Virtual Environment), which helps us to communicate and spread information by relaying a message like Chinese whispers. They have introduced pushing mechanism of information, a compact comment structure, a hierarchical category and visualization of information path to simplify the exchange process.
To control the above preprocessing and information gathering process, they assign a weight to information delivery and a preprocessing process. In a preprocessing stage, they made an experiment to weight examples, and discovered a rule supported by many examples. For time-series data they developed a preprocessing to infer missing data and common periods, based on which a simple mining system discovers simple and easy rules.
The investigators evaluated the system based on some real data sets and compare its results with ACCESS : a popular database software, and found that the proposed system processes the some case with fewer steps.
|
