決策樹形式知識之線上預測系統架構

dc.contributor.author馬芳資zh_tw
dc.contributor.author林我聰zh_tw
dc.contributor.authorFang-Tz Maen_US
dc.contributor.authorWoo-Tsong Linen_US
dc.date.accessioned2014-10-27T15:18:36Z
dc.date.available2014-10-27T15:18:36Z
dc.date.issued2003-10-??zh_TW
dc.description.abstract本研究提出一個決策樹形式知識的線上預測系統架構,其主要的目在於提供一個Web Based的知識發掘(Knowledge Discovery, KD)及線上預測系統,而我們藉由使用這個系統可以進行歸納學習出決策樹形式的知識,並且在線上使用決策樹的知識來做分類和預測的工作。它的組成元件包含三個子系統:知識學習子系統、合併選擇決策樹子系統、線上預測子系統;三個儲存庫:決策樹知識法則庫、例子資料庫、和歷史知識法則庫;以及三個導入知識法則的介面:上傳例子集資料介面、輸入決策樹知識法則介面、及轉換決策樹PMML(Predictive Model Markup Language)文件模組等。就整體系統運作流程而言,在知識學習方面,我們首先上傳例子集,接著使用知識學習子系統來發掘出知識,然後直接儲存於知識法則庫內。而在知識使用方面,我們可以利用線上預測子系統來存取知識法則庫內的知識以進行分類和預測的工作。在知識溝通方面,本系統提供一個轉換PMML格式文件的模組,方便導入其他採礦工具所歸納學習出之決策樹形式的知識。而在知識整合方面,本系統使用合併選擇決策樹子系統來合併多棵決策樹形式的知識而成一棵決策樹。運用這個子系統有助於維護決策樹法則知識庫內的知識,而讓決策樹形式的知識在保有簡單樹狀結構下,進行知識法則的擴充,並且簡單樹狀結構有助於線上預測子系統對於系統預測結果之解釋和說明。有關後續研究方面,本研究擬實作此架構的元件,且對於合併決策樹方面,提出一些修剪策略來提昇決策樹之預測準確度,以及如何有效維護決策樹知識法則庫內的知識等課題。zh_tw
dc.description.abstractThis paper presents an on-line decision tree based predictive system architecture. The architecture contains nine components, including a database of the examples, a learning system of the decision trees, a knowledge base, a historical knowledge base, a maintaining interface of the decision trees, an interface to upload training and testing examples, a PMML (Predictive Model Markup Language) translator, an on-line predictive system, and a merging optional decision trees system. There are three channels to import knowledge in the architecture; the developers can upload the examples to the learning system to induce the decision tree, directly input the information of decision trees through the user interface, or import the decision trees in PMML format. In order to integrate the knowledge of the decision trees, we added the merging optional decision trees system into this architecture. The merging optional decision trees system can combine multiple decision trees into a single decision tree to integrate the knowledge of the trees. In the future research, we will implement this architecture as a real system in the web-based platform to do some empirical analyses. And in order to improve the performance of the merging decision trees, we will also develop some pruning strategies in the merging optional decision trees system.en_US
dc.identifier2F3633E4-F352-BC7F-0487-35237F4CAC59zh_TW
dc.identifier.urihttp://rportal.lib.ntnu.edu.tw/handle/20.500.12235/15157
dc.language中文zh_TW
dc.publisher國立台灣師範大學圖書資訊研究所zh_tw
dc.publisherGraduate Institute of Library and Information Studiesen_US
dc.relation29(2),60-76zh_TW
dc.relation.ispartof圖書館學與資訊科學zh_tw
dc.subject.other決策樹zh_tw
dc.subject.other知識整合zh_tw
dc.subject.other線上預測系統架構zh_tw
dc.subject.otherDecision treeen_US
dc.subject.otherKnowledge integrationen_US
dc.subject.otherOn-Line predictive system architectureen_US
dc.title決策樹形式知識之線上預測系統架構zh-tw
dc.title.alternativeAn On-Line Decision Tree-Based Predictive System Architecturezh_tw

Files

Original bundle

Now showing 1 - 1 of 1
No Thumbnail Available
Name:
ntnulib_ja_A1201_2902_060.pdf
Size:
220.38 KB
Format:
Adobe Portable Document Format