When to Update Sequential Patterns in Stream Data?

Title: When to Update Sequential Patterns in Stream Data?

Abstract: This research investigates the optimal time to update sequential patterns in stream data. It proposes a method called TPD (Tradeoff between Performance and Difference) to determine when to update sequential patterns by making a tradeoff between the performance of increasingly updating algorithms and the difference of sequential patterns. The study uses experimental data from GSM alarm data sets to verify the proposed method.

Main Research Question: How can we determine the optimal time to update sequential patterns in stream data while balancing the performance of increasingly updating algorithms and the difference of sequential patterns?

Methodology: The study first defines a distance measure between old and new sequential patterns, which is proven to be a distance. It then proposes the TPD method to decide when to update sequential patterns by making a tradeoff between the performance of increasingly updating algorithms and the difference of sequential patterns.

Results: Experiments on two GSM alarm data sets show that, as the size of incremental windows grows, the values of the speedup (performance) and the values of the difference (between old and new sequential patterns) decrease and increase respectively. The study also finds that the incremental ratio determined by the TPD method does not monotonically increase or decrease but changes in a range between 18 and 30 percentage for the IUS algorithm.

Implications: The TPD method provides a practical solution to the problem of determining when to update sequential patterns in stream data. It helps balance the performance of increasingly updating algorithms and the difference of sequential patterns, ensuring that updates are made when necessary without unnecessarily burdening the system with frequent updates.

Link to Article: https://arxiv.org/abs/0203028v2 Authors: arXiv ID: 0203028v2