THE MULTIPLE REGRESSION MODEL WITH DICHOTOMOUS VARIABLES IN ANALYSIS OF MULTI-SECTIONAL DEMAND FOR CONNECTIVITY SERVICES – APPROACH BASED ON PER SECOND BILLING
THE MULTIPLE REGRESSION MODEL WITH DICHOTOMOUS VARIABLES IN ANALYSIS OF MULTI-SECTIONAL DEMAND FOR CONNECTIVITY SERVICES – APPROACH BASED ON PER SECOND BILLING

Summary/Abstract: The article presents the results of comparative research of the effectiveness of two types of models in terms of approximation and short-term forecasting of the multi-sectional demand for connectivity services. The presented results of the analyses are related to the selection of an appropriate forecasting method as an element of the Prediction System dedicated to telecommunications operators. The first tested model was a multiple regression model with dichotomous explanatory variables. The second model was a multiple regression model with dichotomous explanatory variables and autoregression. In both models, the dependent variable was the hourly counted seconds of outgoing calls within the network of the selected operator. Telephone calls were analysed in terms of such classification factors as: type of day, category of call, group of subscribers. Taking into account all levels of classification factors of the explanatory variable, 35 dichotomous explanatory variables were specified. The defined set of dichotomous explanatory variables was used in the estimation process of both compared regression models. However, in the second model, first-order autoregression was additionally applied. The second model (multiple regression model with dichotomous explanatory variables with first-order autoregression) was found to have higher approximation and predictive capabilities than the first model (multiple regression model with dichotomous explanatory variables without autoregression).

• Details
• Contents