Methods for predicting traffic to an organization's web resources

With the development of technology, the volume of online commerce is increasing, highlighting the need for analysis and forecasting of the traffic to informational web resources. The article examines various tools for collecting and analyzing web traffic, such as Google Analytics, Yandex.Metrica, Adobe Analysis, LiveInternet, and ClickTale. Based on the analysis of the data obtainable from these sources, key metrics for forecasting traffic were identified. The metrics were grouped by qualitative and quantitative characteristics. Quantitative metrics were used for forecasting with statistical models, while qualitative metrics served as descriptive fields for the series forecasted using machine learning. For example, traffic, views, and depth of visit were identified as quantitative variables, while events, sessions, bounce rate, goals, conversions, traffic sources, and returns were classified as qualitative. Additionally, both statistical models for forecasting web traffic and models using machine learning were constructed and analyzed. The models used included SARIMA (Seasonal Autoregressive Integrated Moving Average), the Holt–Winters model, and an autoregressive model with gradient boosting. This analysis allowed for a comparison of the accuracy of the models using criteria such as mean squared error, root mean squared error, and mean absolute error. In conclusion, the gradient boosting model demonstrated the lowest average error rates, indicating the highest accuracy; however, for some series, statistical models showed greater accuracy, suggesting the need for a combination of models to achieve the most optimal results.