| Resumo: | The ever increasing usage of networks around the world made the telecommunication companies to start planning ahead out of necessity. The present work is focused on analysing and understanding which of the tested predictive models best suits Long Term Evolution (LTE) behaviour regarding its capacity, by forecasting several Key Performance Indicators (KPI) originated from network daily cells and dedicated to the same subject. Many were the tested models, ranging from the benchmark models (which comprise naïve, seasonal naïve and drift), to Exponential Smoothing (ES), AutoRegressive Integrated Moving Average (ARIMA), Theta and Linear Regression and also including models used in the latest M4 competition. The inherent purpose was not to find a model that was definitely better than the remaining, but instead to understand which model can best serve the KPI under analysis and the predicted forecasted horizon. The present study forecasts and analyses several different models in order to achieve better predictive results so that telecommunication companies can make more informed decisions regarding network planning.
|
|---|