On-demand DWDM design using machine learning

This paper demonstrates ML techniques and regression modeling to predict the distribution of optical performance-dependent parameters (PDPs) and performance monitoring factors (PMFs) of the unestablished lightpaths. Specifically, we discuss and evaluate the performance of the ML approaches such as SVM, KNN, and LD-PCA by utilizing raw data obtained from the optical performance monitoring tools such as wavelength division multiplexer (WDM) analyzer and optical spectrum analyzer. It is necessary to guarantee the quality of transmission (QoT) in fiber optic communication to meet the end-user requirements. Conventional approaches acquire more complex system designs to ensure the QoT that leads to computational difficulties. ML has accomplished haulage in the lightpath community in recent years to achieve high-level accurate QoT. In this work, we evaluate the performance of the proposed ML approaches on fiber networks in terms of fit-model analysis. Here, we also compared the performance of the conventional and predicted model by attributing synthetic data such as R -value, marginal error calculations, A.D factor, and standard deviations, which are retrieved from the ML regression approaches. In addition, we also provide an ML-based analytical model, which detects, identifies the quality of PMFs, allows performing decision making in unestablished lightpath implementation, and meets the state-of-art QoT classification models. Graphical abstract