Abstract

The article addresses the importance of effective planning in both state and private companies to forecast their future economic activity. It mentions various mathematical models, including linear and curvilinear ones, used to create reliable projections. The State Agency for Regulation and Control of Telecommunications of Ecuador (ARCOTEL) monitors diverse variables, such as the number of users in internet service providers, using curvilinear regressions to obtain approximations when the exact functional form of the regression curve is unknown. The study's methodology includes data pre-processing, homogeneity analysis, and data normalization. Statistical tests like the Mann-Kendall Test and the Helmert statistical test were used to assess trends and structural changes in the time series. Subsequently, the data underwent polynomial fitting, ranging from a linear fit to a polynomial of degree 6. The results show that this polynomial fit provides an adequate representation of the time series, with high correlation coefficients indicating a good model fit. The percentage, absolute, and mean square errors are relatively low, suggesting acceptable precision and agreement between the fitted values and the actual values.