Utilizing An Ensemble STL Decomposition And GRU Model For Base Station Traffic Forecasting

Owing to the increasing sophistication,volume,and scale of the demands of rapidly developing wireless communications networks and systems,corresponding network traffic volume has similarly scaled in both proportion and complexity.Base station traffic forecasting is at the frontline of research currently as a response to meet the needs of future users.The author proposes an ensemble model that combines time series decomposition via seasonal and trend decomposition using loess(STL),individually forecasting decomposed components with gated recurrent unit(GRU)neural network model,then recombining forecasts.The rationale behind the hybrid model is that the decomposition reduces the effect of noise and outliers in the time series data,thereby enabling improved results when forecasting traffic data when compared against standalone statistical or machine learning techniques.The study found some success in increasing performance gains when juxtaposed against comparable popular contemporaneous models.The results have shown modest improvements to forecast accuracy,thereby implying performance gains in traffic prediction.The model yielded an average of 23%increase in predictive accuracy when compared against standard GRU and modern time series analysis techniques in relation to forecasting base station traffic data.The key contributions of this study are:1.the formulation of a viable model for base station traffic prediction;2.the layout for the theoretical basis of model performance;and3.the evidence-based proof of increased model performance against state of the art models.These contributions are further supported by further discussion of model use cases and possible future research directions.