We apply Quantile Regression (QR) for lightpath quality-of-transmission (QoT) estimation with the aim of identifying uncertain decisions and then exploit Shapley Additive Explanations (SHAP) to quantify lightpath features’ importance by means of SHAP values and validate the model’s decisions in a post-processing phase. Numerical results show that our approach can eliminate more than 98% of false predictions and that SHAP values can validate up to 90% of the model's uncertain decisions.
Using SHAP Values to Validate Model’s Uncertain Decision for ML-based Lightpath Quality-of-Transmission Estimation / Houssiany, Hadi; Ayoub, Omran; Rottondi, Cristina; Bianco, Andrea. - ELETTRONICO. - (2023), pp. 1-5. (Intervento presentato al convegno 23rd International Conference on Transparent Optical Networks (ICTON) tenutosi a Bucharest, Romania nel 2-6 July 2023) [10.1109/ICTON59386.2023.10207178].
Using SHAP Values to Validate Model’s Uncertain Decision for ML-based Lightpath Quality-of-Transmission Estimation
Rottondi, Cristina;Bianco, Andrea
2023
Abstract
We apply Quantile Regression (QR) for lightpath quality-of-transmission (QoT) estimation with the aim of identifying uncertain decisions and then exploit Shapley Additive Explanations (SHAP) to quantify lightpath features’ importance by means of SHAP values and validate the model’s decisions in a post-processing phase. Numerical results show that our approach can eliminate more than 98% of false predictions and that SHAP values can validate up to 90% of the model's uncertain decisions.File | Dimensione | Formato | |
---|---|---|---|
ICTON_reference.docx (1).pdf
accesso aperto
Tipologia:
2. Post-print / Author's Accepted Manuscript
Licenza:
PUBBLICO - Tutti i diritti riservati
Dimensione
516.38 kB
Formato
Adobe PDF
|
516.38 kB | Adobe PDF | Visualizza/Apri |
Using_SHAP_Values_to_Validate_Models_Uncertain_Decision_for_ML-based_Lightpath_Quality-of-Transmission_Estimation.pdf
non disponibili
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Non Pubblico - Accesso privato/ristretto
Dimensione
1.19 MB
Formato
Adobe PDF
|
1.19 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/2981205