| Item type |
紀要論文 / Departmental Bulletin Paper(1) |
| 公開日 |
2020-11-24 |
| タイトル |
|
|
タイトル |
背景情報を活用するデータマイニングシステムの開発 |
|
言語 |
ja |
| タイトル |
|
|
タイトル |
Development of Data Mining System using Background Information |
|
言語 |
en |
| 言語 |
|
|
言語 |
jpn |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
data mining |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
time-series data |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
annotation |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
dynamic time warping |
| キーワード |
|
|
言語 |
en |
|
主題Scheme |
Other |
|
主題 |
decision tree learning |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_6501 |
|
資源タイプ |
departmental bulletin paper |
| ID登録 |
|
|
ID登録 |
10.34411/00001093 |
|
ID登録タイプ |
JaLC |
| 著者 |
杉村, 博
松本, 一教
Sugimura, Hiroshi
Matsumoto, Kazunori
|
| 抄録 |
|
|
内容記述タイプ |
Abstract |
|
内容記述 |
This paper proposes a system which datamines annotated time series leading by a discovery of feature patterns. Several studies propose extraction methods from time series data which is sequence of numerical values. They generally have a lot of important information in background, but it is not included in data. We point out that analysis methods without background information have limitations. We therefore develop two mechanisms that uses annotations which are compact expressions of back ground informaion. First mechanism runs in two stages. In the first stage, the system discovers important feature patterns. For this purpose, we propose a feature importance measure which is called FI. The second stage builds IF-THEN rules that predict future behaviors based on the annotations. For the understandability of the prediction rule, we use the IF-THEN rule style that represents X→ Y as the association rule. In order to extract the rule from time series data, we propose a new datamining method. The second mechanism is an automation of the annotation process. An automation function and/or decentralization function for this task is needed because human annotations require high costs. We thus also develop an automatic annotation method for financial time series data based on web news sites. We explain how these two mechanisms are harmonized in the entire process. |
|
言語 |
en |
| 書誌情報 |
神奈川工科大学研究報告.B,理工学編
巻 37,
p. 65-70,
発行日 2013-03-20
|
| 出版者 |
|
|
出版者 |
神奈川工科大学 |
| ISSN |
|
|
収録物識別子タイプ |
PISSN |
|
収録物識別子 |
09161902 |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AN10074179 |
| フォーマット |
|
|
内容記述タイプ |
Other |
|
内容記述 |
application/pdf |
| 著者版フラグ |
|
|
出版タイプ |
VoR |
|
出版タイプResource |
http://purl.org/coar/version/c_970fb48d4fbd8a85 |
kkb-037-013.pdf (3.9 MB)
