Real and Synthetic Dataset for Slice Performance Forecasting

This dataset combines both synthetic and real-world data. The synthetic component was generated from simulated data that underwent thorough analysis to identify patterns of seasonality and correlations among variables. In addition, real data series—collected from operational 5G network environments—are included to capture actual slice performance behavior. Its primary aim is to support the training and evaluation of various forecasting strategies for slice performance in 5G networks, as detailed in the article: "A Comparative Study of Training Strategies for Slice Performance Forecasting in B5G Networks". The dataset comprises synthetic canonical series that accurately mirror the characteristics of the original data, alongside real-world series representing genuine network conditions. All series are stored in Parquet format, facilitating efficient access. They are specifically designed to aid in the comparison and assessment of different forecasting methods within the context of 5G and B5G networks.