David Lo Siau-Cheng Khoo Chao Liu DASFAA 2008 Efficient Mining of Recurrent Rules from a Sequence Database 1

Outline 2 Introduction Preliminaries Generation of Recurrent Rules Algorithm Performance Evaluation Conclusion

Introduction 3 Mining for knowledge from data has been shown useful for many purposes ranging from finance, advertising, bio- informatics and recently software engineering. “ Whenever a series of precedent events occurs, eventually another series of consequent events occurs ”

Introduction 4 1. Resource Locking Protocol: Whenever a lock is acquired, eventually it is released. 2. Internet Banking: Whenever a connection to a bank server is made and an authentication is completed and money transfer command is issued, eventually money is transferred and a receipt is displayed. 3. Network Protocol: Whenever an HDLC connection is made and an acknowledgement is received, eventually a disconnection message is sent and an acknowledgement is received.

Preliminaries 5 Linear-time Temporal Logic (LTL) ‘G’ specifies that globally at every point in time a certain property holds. ‘F’ specifies that a property holds either at that point in time or finally (eventually) it holds. ‘X’ specifies that a property holds at the next event.

Preliminaries 6

7 Checking/Verifying LTL Expressions. (main, lock) → (unlock, end) (main, lock, use) → (unlock, end)

Generation of Recurrent Rules 8 Concepts and Definitions

Generation of Recurrent Rules 9 Concepts and Definitions

Generation of Recurrent Rules 10 Concepts and Definitions

Generation of Recurrent Rules 11 Concepts and Definitions

Generation of Recurrent Rules 12 Concepts and Definitions

Generation of Recurrent Rules 13 Concepts and Definitions s-support:2 i-support:2 confidence:1

Generation of Recurrent Rules 14 Concepts and Definitions

Generation of Recurrent Rules 15 Concepts and Definitions s-support:2 i-support:2 confidence:1

Generation of Recurrent Rules 16 Apriori Properties and Non-redundancy

Generation of Recurrent Rules 17 Apriori Properties and Non-redundancy

Generation of Recurrent Rules 18 Apriori Properties and Non-redundancy

Algorithm 19

Performance Evaluation 20

Conclusion 21