A New Deep Learning Hybrid Model for Accurate Web Traffic Time Series Forecasting

Time series forecasting is a critical task in diverse domains, including finance, marketing, and the analysis of web traffic. Web-based services heavily rely on accurate short- and long-term traffic predictions to ensure seamless operations, optimal resource allocation, and efficient handling of server outages. This paper presents the combination of statistical and machine learning techniques to improve the forecasting accuracy of time series data. In particular, the Seasonal Autoregressive Integrated Moving Average (SARIMA) model is combined with the Informer deep learning model, for the first time, to forecast web traffic pages' view. In the proposed method, the data is initially decomposed into trend, seasonality, and residuals. The seasonality and residuals are then forecasted using the SARIMA technique, while the non-linear trend is forecasted using the Informer deep learning model. Results indicates that the proposed hybrid SARIMA-Informer model yields superior accuracy in web traffic time series forecasting compared to other techniques. This marks a major advancement in the area of time series prediction for web traffic forecasting.