Events
Your location now is: Home > Lectures > Content

(September 26)Summarizing Uncertain Transaction Databases by Probabilistic Tiles
Sep 22, 2016

Title:

Summarizing  Uncertain Transaction Databases by Probabilistic Tiles

Lecturer:

Dr. Ling CHEN

Time:

10:00, September 26,  2016

Venue:

305, Section  IV, Main Building, North Campus

Lecturer  Profile

Dr.  Ling Chen is a senior lecturer with the Centre for Quantum Computation and  Intelligent Systems, University of Technology, Sydney. She received Ph.D. from  Nanyang Technological University, Singapore. Before joining UTS, she was a  postdoc research fellow with L3S Research Center, University of Hannover,  Germany. Her research interests include data mining and machine learning,  social network analysis and recommender systems. She has published more than 50  papers in major conferences and journals including SIGKDD, ICDM, SDM, WWW and  ACM TOIS.

Lecture  Abstract

Abstract:  Transaction data mining is ubiquitous in various domains and has been  researched extensively. In recent years, observing that uncertainty is inherent  in many real world applications, uncertain data mining has attracted much  research attention. Among the research problems, summarization is important  because it produces concise and informative results, which facilitates further  analysis. However, there are few works exploring how to effectively summarize  uncertain transaction data. In this work, we formulate the problem of  summarizing uncertain transaction data as Minimal Probabilistic Tile Cover  Mining, which aims to find a high-quality probabilistic tile set covering an  uncertain database with minimal cost. We define the concept of Probabilistic  Price and Probabilistic Price Order to evaluate and compare the quality of  tiles, and propose a framework to discover the minimal probabilistic tile  cover. The bottleneck is to check whether a tile is better than another  according to the Probabilistic Price Order, which involves the computation of a  joint probability. We prove that it can be decomposed into independent terms  and calculated efficiently. Several optimization techniques are devised to  further improve the performance. Experimental results on real world datasets  demonstrate the conciseness of the produced tiles and the effectiveness and  efficiency of our approach.

Previous:(September 27)Shannon Meets Huffman – Huffman Coded Spatial Modulation
Next:(September 24)Algebraic Synthesis of Mechanisms and Robots

South Campus
Add: 266 Xinglong Section of Xifeng Road, Xi’an, Shaanxi 710126
Tel: 86-29-81891818
North Campus
Add: No. 2 South Taibai Road, Xi’an, Shaanxi 710071
Tel: 86-29-88202212