Deep Learning and Heuristic Approach Assisted Dynamic Spectrum Assignment with Fragmentation Minimization in EON-SDM

Dynamic Core Allocation with Fragmentation Optimization is a vital aspect of spectrum management in contemporary communication networks. This paper introduces a new method that utilizes machine learning techniques, specifically deep learning with TensorFlow and Keras, to enhance core allocation and reduce fragmentation in the network. Weemploythe Dijkstra algorithm to ascertain the k shortest routes connecting a designated source and destination. Next, we calculate fragmentation coefficients using the Contiguous Aligned Slot Ratio (CASR) measure. Eventually, we utilize a deep learning model that has been trained using these fragmentation coefficients to forecast whether a core will be accepted or rejected, thus facilitating the allocation of cores in a dynamic manner. In addition, we present a function that calculates the Bandwidth Blocking Probability (BBP) in order to assess the effectiveness of the allocation technique. The findings suggest that our technique is effective in minimizing fragmentation and boosting spectrum utilization, resulting in lower BBP values and improved network performance.