Analysis and Mitigation of Unwanted Biases in ML-based QoT Classification Tasks

We address the problem of mitigating biases in models used for the quality of transmission prediction. The proposed method reduces the relative accuracy difference between samples with different feature values by up to 45%.