Download presentation
Presentation is loading. Please wait.
Published byFrancis Stukey Modified over 9 years ago
1
Rule Discovery from Time Series Presented by: Murali K. Kadimi
2
What’s Next 4 Goal of the Paper 4 Time-Series Discretization 4 Rule Discovery from Sequences 4 Experimental Results 4 Future Work and Conclusions
3
Authors’ Goal 4 Extraction of Patterns Step1: Discretization 4 Rule Discovery from Patterns Step2: Rule Discovery
4
Authors’ Goal (contd....) Examples: Web Mining : - 60% of clients who placed an online order in /company/products/product1.html, also placed an online order in /company1/products/product4 within 15 days. Telecom Mining: - link Alarm, link failure (5) (60) High Fault Rate (0.7)
5
Discretization
6
Discretization (contd...) 4 Sequence ( s ) s = (x 1, …, x n ) 4 Window Size ( w ) 4 Subsequences of Data ( s i ) s i = (x i, …, x i+w-1 )
7
Discretization (contd...)
8
4 Original time series = (1,2,1,2,1,2,3,2,3,4,3,4) 4 Window width = 3 4 Discretized series (D(s)) = (a1,a2,a1,a2,a3,a1,a2,a3,a1,a2)
9
Discretization (contd...) 4 Clustering - A Greedy algorithm - k-means algorithm 4 Choice of Parameters w, v, d, k
10
Discretization (contd...)
11
Rule Discovery Rule Format T A B 4 Frequency F(A) Confidence T F(A,B,T) c(A B) = -------------- F(A) F(A,B,T) = |{i|a i = A B {a i+1, …, a i+T-1 }}|
12
Rule Discovery (contd...) 4 Rule Generation AprioriSome, AprioriAll, DynamicSome 4 Informative Rules J-measure 4 Multiple Time Series
13
Some Practicals 4 Data-Sets Stock data Rules Discovered 20 s18 s4 Sup(2.8%) Conf(59.6%) 4 Robustness
14
Some Practicals (contd...) TASA Aim Two Central Phases Pattern Discovery Pattern Presentation Example Rules 5,60 A,B C (0.7)
15
What it Means 4 The proposed technique is just an Exploratory method.
16
Next Steps 4 Computational costs Vs Quality of Rules 4 Search Algorithm to retrieve Informative Rules
17
Important Links 4 Das, Lin, Mannila, Renganathan, and Smyth, "Rule Discovery from Time Series", KDD, 16-22, 1998.Rule Discovery from Time Series 4 Agarwal and Srikant, “Mining Sequential Patterns”, ICDE’95.Mining Sequential Patterns 4 TASA: Telecommunication Alarm Sequence Analyzer.Telecommunication Alarm Sequence Analyzer 4 Software for Discovery of Sequential Patterns http://www.channel1.com/users/gps1/software/se quence.html http://www.channel1.com/users/gps1/software/se quence.html
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.