Intelligent Search for Correlated Alarms in Noisy Databases

Title: Intelligent Search for Correlated Alarms in Noisy Databases

Abstract: This research focuses on discovering alarm correlation rules from databases containing noise data. The study defines two parameters, Win_freq and Win_add, to measure noise data and presents the Robust_search algorithm to solve the problem. Experiments with alarm data containing noise show that the Robust_search Algorithm can discover more rules with a bigger size of Win_add. The research also compares two interestingness measures, confidence, and correlation.

Main Research Question: How can we develop an algorithm to discover correlated alarms in databases containing noise data?

Methodology:

1. Definition of Noise Data: The study defines noise data as irrelevant or misleading information that can interfere with the discovery of genuine alarm correlations.

2. Robust_search Algorithm: The algorithm is designed to search for correlated alarms in databases containing noise data. It uses the Win_freq and Win_add parameters to measure noise data and adjusts the search parameters accordingly.

Results:

1. Experiments: The study conducts experiments with alarm data containing noise, demonstrating that the Robust_search Algorithm can discover more rules with a bigger size of Win_add.

2. Comparison of Interestingness Measures: The research compares two interestingness measures, confidence, and correlation, finding that both can be useful in different contexts.

Implications:

1. The study provides a method for discovering correlated alarms in databases containing noise data, which is crucial for improving the reliability and service of modern telecommunications networks.

2. The comparison of interestingness measures offers insights into the best approach for different situations, helping users make informed decisions about the data analysis process.

Link to Article: https://arxiv.org/abs/0109042v1 Authors: arXiv ID: 0109042v1