Knowledge Engineering & Artificial Intelligence Lab (知識工程與人工智慧)

Presentation on theme: "Knowledge Engineering & Artificial Intelligence Lab (知識工程與人工智慧)"— Presentation transcript:

1 Knowledge Engineering & Artificial Intelligence Lab (知識工程與人工智慧)
Advisor: Tzung-Pei Hong (洪宗貝) Presenter: Chun-Wei Lin (林浚瑋) Postdoc research fellow (博士後研究員)

2 Research 近年來，我們對於人工智慧、軟式計算及資料 探勘相關技術的研究已有豐富成果，其中對於 類神經網路、模糊理論及遺傳演算法等研究主 題除了理論方法上的創新與改進之外，並成功 地將之應用於工作排程、腦瘤診斷、遊戲樹、 網際網路等實際問題，且已發表二百餘篇論文 於多種國際著名期刊及會議。除此之外，我們 也執行過多個與上述領域方面相關的國科會計 畫，獲得相當不錯之經驗及成果。

3 Study topics (1) Soft Computing Fuzzy Set Neural Network
Genetic Algorithm Rough Set Simulated Annealing Ant Systems Probabilistic Reasoning Fuzzy Measure Hybrid Systems

4 Study topics (2) Artificial Intelligence Expert Systems
Machine Learning Knowledge Integration Heuristic Search AI in Games

5 Study topics (3) Uncertain Scheduling Data Mining
Minimum completion time Data Mining Association rules mining Fuzzy data mining / Quantitative data mining Sequential patterns Incremental mining Online data mining Privacy preserving data mining Knowledge integration

6 Study topics (4) Object-Oriented Data Warehousing
Uncompressed model Compressed Model Hybrid Model Management Information Systems Steel Company Campus Computerization Nursing Institute Electronic-case System

7 Study topics (5)

8 NSC projects 24 NSC projects In 5 years 計畫名稱 起訖年月 擔任工作 經費總額
隱私防護資料探勘新方法之研究( E MY3) 2011/8/1~2014/7/31 主持人 1,970,000 加速不確定性與模糊資料探勘之研究( E ) 2010/8/1~2011/7/31 587,000 彩色三維文物模型壓縮與保存技術之研究 －以高雄市立歷史博物館數位典藏為例 ( H ) 共同主持人 621,000 整合知識本體於蒙地卡羅樹搜尋與其應用 ( E MY3) 2010/1/1~2012/12/31 7,181,000 彩色三維文物模型之壓縮與保存技術之研究 －以高雄市立歷史博物館數位典藏為例( H ) 2009/8/1~2010/7/31 902,000 模糊約略集合論於資料挖掘之應用( E ) 智慧型網頁資訊擷取與融合之研究( E ) 2008/8/1~2009/7/31 －以高雄市立歷史博物館數位典藏為例( H ) 2008/3/1~2009/7/31 931,000 語意網路本體知識選擇、建立及整合技術之研究(3/3) ( E ) 2007/8/1~2008/7/31 879,000 下一世代資訊通訊網路尖端技術與應用(二)－ 子計畫五：網路安全(4/4)( E PAE) 2007/4/1~2008/7/31 6,321,000

9 Data mining scenario Supermarket Products Relationship between items
How to arrange products in the supermarket? Products Jerry Relationship between items

10 Knowledge and strategy
The role of data mining Data Information Useful patterns Transaction data Data Mining Knowledge and strategy Preprocess data

11 Association rules An example TID Items T1
milk, bread, cookies, beverage T2 milk, bread, cookies T3 bread, cookies, beverage T4 milk, bread T5 Milk Bread minsup IF bread is bought THEN milk is bought minconf

12 Real-world applications
Dynamic databases Insertion Deletion Modification New records TID Items 5 A, D, E 6 B, C Insertion Original database Large 1-itemsets Itemset Support {A} 2 {B} 3 {C} {E} TID Items 1 A, C, D 2 B, C, E 3 A, B, C, E 4 B, E Large 2-itemsets Itemset Support {AC} 2 {BC} {BE} 3 {CE} Large 3-itemsets Itemset Support {BCE} 2

13 Architecture

14 Knowledge from different sources
… Mined Knowledge 1 Mined Knowledge 2 Mined Knowledge 3 Mined Knowledge n Which is correct? ? ? ? Decision Maker

15 知識萃取技術與應用之整合研究 應用層 探勘知識整合技術之研究 知識萃取層 使用者介面層 企業行銷知識之萃取與應用
子計畫七 探勘知識整合技術之研究 本計畫 資料探勘技術於網站評估與效能改善之研究 多質性資料倉儲技術之研究 子計畫二 資料 探勘 與 機器學習技術 前端 處理 技術 後端 智慧型資料探勘前置處理技術之研究 子計畫一 資料輸入 子計畫六 空間與時間資料探勘技術之研究 子計畫三 跨影像與文字資料探勘技術之研究 子計畫四 應用層 知識萃取層 使用者介面層 知識萃取技術與應用之整合研究

16 Knowledge Integration
Transaction database 1 Transaction database 2 Transaction database 3 Transaction database n … … Data Mining System 1 Data Mining System 2 Data Mining System 3 Data Mining System n … Mined Knowledge Mined Knowledge Mined Knowledge Mined Knowledge Local ItemSets Local ItemSets Local ItemSets Local ItemSets Global Item Set Encoding Ontology Intermediary Representation Intermediary Representation Intermediary Representation Intermediary Representation ……………… ……………… … Sample Data Knowledge Integration Integrating Integrated Knowledge Base

17 Privacy-Preserving Data Mining
A company B company Sensitive data

18 Privacy-preserving data mining
Insertion deletion Original database Large 1-itemsets Itemset Support {A} 2 {B} 3 {C} {E} TID Items 1 A, C, D 2 B, C, E 3 A, B, C, E 4 B, E Large 2-itemsets Itemset Support {AC} 2 {BC} {BE} 3 {CE} Large 3-itemsets Itemset Support {BCE} 2 TID Items 5 C, D, E 6 A, C Hiding the sensitive rules

19 Thank you for listening