Pre-, In-, and Post-Processing Class Imbalance Mitigation Techniques for Failure Detection in Optical Networks

TL;DR

This paper evaluates various class imbalance mitigation techniques for failure detection in optical networks, comparing their effectiveness and efficiency to guide practical deployment choices.

Contribution

It provides a comparative analysis of pre-, in-, and post-processing methods, identifying the most effective and efficient techniques for optical network failure detection.

Findings

Threshold Adjustment yields the highest F1 score improvement (15.3%).

Random Under-sampling offers the fastest inference speed.

Trade-offs between performance gains and computational complexity are highlighted.

Abstract

We compare pre-, in-, and post-processing techniques for class imbalance mitigation in optical network failure detection. Threshold Adjustment achieves the highest F1 gain (15.3%), while Random Under-sampling (RUS) offers the fastest inference, highlighting a key performance-complexity trade-off.