Multimodal Deep Learning-based Demand Forecasting in Network Slicing

One of the critical benefits of emerging wireless networks is the provision of accurate demand predictions. Resource availability is one of the essential factors ensuring such connectivity in heterogeneous networks. Despite extensive research interest in this domain, the fundamental issues are to ensure efficient allocation and exploitation of network resources. This paper proposes a multi-model-based model to forecast demand requirements utilizing deep learning techniques in network slicing. We present a framework that employs multiple forecasting models to perform forecasting by using historical information and cognitively selecting the most accurate forecasting model. Furthermore, we conduct a detailed analysis of several forecasting models from various papers. According to the findings, the proposed forecasting framework favors deep learning models and enhances fairness and guarantees experience quality. Moreover, we have demonstrated that the suggested approach can account for forecasting variations.