Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

Multi-focal Learning and Its Application to Customer Service Support

Presenter : Tsai Tzung Ruei　Authors : Yong Ge, Hui Xiong, Wenjun Zhou, Ramendra Sahoo,Xiaofeng Gao,Weili Wu

2009.SIGKDD

國立雲林科技大學National Yunlin University of Science and Technology

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Outline

Motivation Objective Methodology Experiments Conclusion Discussions Comments

2

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Motivation

All the problem descriptions for the same problem are provided by customers with diverse background and these problem descriptions can be quite different.

3

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Objective

To formalize a multi-focal learning problem, where training data are partitioned into several different focal groups and the prediction model will be learned within each focal group.

4

focal groups

Problem descriptions Problem Solution

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology(1/3)

5

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology(2/3)

Focal Group Formation:CORRELATION

Focal Group Formation: ONTOLOGY

6

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology(3/3)

Risk Analysis of Multi-Focal Learning

7

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments(1/5)

Results on Problem Logs Performance Comparisons Results on Synthetic Data Case Study

8

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments(2/5)

Results on Problem Logs

9

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments(3/5)

Performance Comparisons

10

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments(4/5)

Results on Synthetic Data

11

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments(5/5)

Case Study

12

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Conclusion

The multi-focal learning allows the learning algorithms to mitigate the influence of the diversities inherent in training data, and thus leads to better learning performances.

Experimental results show that both CORRELATION and ONTOLOGY have led to better learning performances than other focal-group formation methods, such as the methods based on clustering and random-partition, while the learning performance by ONTOLOGY is lightly better than that by CORRELATION.

13

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Discussions

For instance, let us consider a video surveillance system. There are different types of moving objects, such as cars, bikes, and human beings. Those moving objects have different sizes, speed, and moving capabilities. To better capture abnormal moving patterns, it is expected to apply the multi-focal learning techniques to first group moving objects into different focal groups. The detection of abnormal moving patterns can then be performed within different focal groups.

14

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Comments

Advantage To boost the learning accuracies of existing learning algorithms, such as

Support Vector Machines (SVMs), for classifying customer problems.

Drawback Some mistakes

Application Customer Service Support

15