advertisement

An Extended Two-Phase Architecture for Mining Time Series Data

67 %
33 %
advertisement
Information about An Extended Two-Phase Architecture for Mining Time Series Data

Published on February 4, 2008

Author: sosorry

Source: slideshare.net

advertisement

An Extended Two-Phase Architecture for Mining Time Series Data An-Pin, Chen; Yi-Chang, Chang; Nai-Wen, Hsu {apc, ycchen, sosorry }@iim.nctu.edu.tw 2005 KES@Melbourne

APC Laboratory Financial Investment Decision Intelligent Decision Support System Expert System

APC Laboratory

Financial Investment Decision

Intelligent Decision Support System

Expert System

Time series data Vary with time…

Vary with time…

If we could… Forecast Get quantitative result Validate the forecast result

Forecast

Get quantitative result

Validate the forecast result

Outline Literature review Two-phase architecture Experiment result Conclusion

Literature review

Two-phase architecture

Experiment result

Conclusion

Literature review Time series analysis Pattern recognition, Similarity measure Disadvantage Non-quantitative result, Non-statistical meaning Association rule mining Advantage Inference result: (If A  Then B) Disadvantage Either too many or too few rules The wide range rules

Time series analysis

Pattern recognition, Similarity measure

Disadvantage

Non-quantitative result, Non-statistical meaning

Association rule mining

Advantage

Inference result: (If A  Then B)

Disadvantage

Either too many or too few rules

The wide range rules

The Concept of Two-phase architecture How to enforce the effectiveness of rules? Do EDA before association rule mining How to eliminate the low predictability rules? Calculate the accuracy after association rule mining

How to enforce the effectiveness of rules?

Do EDA before association rule mining

How to eliminate the low predictability rules?

Calculate the accuracy after association rule mining

Association rule mining process

Association rule mining process

Association rule mining process

Traditional association rule mining for time series data

Traditional association rule mining for time series data

Traditional association rule mining for time series data

Exploratory data analysis What can EDA do? maximize insight into a data set determine optimal factor settings What does EDA contain?

What can EDA do?

maximize insight into a data set

determine optimal factor settings

What does EDA contain?

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Two-phase architecture

Experiment target Taiwan Weighted Stock Index trading volume Totally 480 days Each day includes 270 ticks

Taiwan Weighted Stock Index trading volume

Totally 480 days

Each day includes 270 ticks

Experiment result & comparison 3 criteria The number of rules The accuracy of rules The effectiveness of rules

3 criteria

The number of rules

The accuracy of rules

The effectiveness of rules

 

 

 

 

 

 

 

 

 

 

Conclusion Two-phase architecture mining is superior than tradition architecture mining by The number of rules The accuracy of rules The effectiveness of rules Some better clustering algorithms may be adopted in the future

Two-phase architecture mining is superior than tradition architecture mining by

The number of rules

The accuracy of rules

The effectiveness of rules

Some better clustering algorithms may be adopted in the future

Thanks for your attention… Q & A

Appendix A : An example of time series data

Raw data in database

Phase 1: Exploratory data analysis (EDA)

Kernel smoothing

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Phase 2: Quantitative inter-transaction association rules

Accuracy Analysis

Accuracy Analysis

Accuracy Analysis

Add a comment

Related pages

LNAI 3681 - An Extended Two-Phase Architecture for Mining ...

An Extended Two-Phase Architecture for Mining Time Series Data 1189 the valid rules would be stored into rule base. In the following experiments, the tradi-
Read more

An Extended Two-Phase Architecture for Mining Time Series ...

Abstract. Time series data vary with time. In the past, most of the researches focused on the matching of feature points or measuring of the similarities.
Read more

An Extended Two-Phase Architecture for Mining Time Series Data

Time series data vary with time. In the past, most of the researches focused on the matching of feature points or measuring of the similarities.
Read more

National Chiao Tung University Institutional Repository:An ...

... a two-phase architecture for mining time series ... An extended two-phase architecture for mining ... architecture for mining time series data is ...
Read more

Time Series Model Query Examples - msdn.microsoft.com

Architects. Overview; ... You add new data to a time series mining model in one of ... When you add new data to a time series model by using a PREDICTION ...
Read more

Dennis Shasha NYU - Time Series in Finance

Time Series in Finance: ... Which research in temporal data mining might help finance? Time series bibliography. ... an extended example.
Read more

Training and Testing Data Sets - msdn.microsoft.com

The information about the size of the training and testing data ... MINING STRUCTURE statement has been extended to ... data from a time series ...
Read more

Mining asynchronous periodic patterns in time series data ...

Mining Asynchronous Periodic Patterns in Time ... Abstract—Periodicy detection in time series data is a challenging ... A two-phase algorithm is ...
Read more