Regression Models for the Analysis of Telecommunications Data in Ecuador

Abstract

The research highlights the importance of mathematical models for better planning in both state and private companies, specifically through curvilinear regressions, to forecast future activities of the State Telecommunications Regulation and Control Agency of Ecuador (ARCOTEL), by analyzing variables such as the number of internet service users.
The study was based on data preprocessing, which included homogeneity analysis and scale changes. Statistical tests such as the Mann-Kendall Test and the Helmert Test were applied to evaluate trends in time series. Subsequently, the data were fitted from a linear model to a polynomial one. Evaluation metrics included absolute, mean, and quadratic percentage errors, as well as coefficients of determination and correlation.
The analysis showed that the sixth-degree polynomial fitting provided an adequate adjustment for the time series, with high correlation coefficients and relatively low absolute and mean percentage errors, suggesting acceptable accuracy between the fitted and actual values. Scaling the data facilitated comparison and analysis, eliminating biases.
The research emphasized the importance of effective planning using mathematical models to predict economic activity in companies. The sixth-degree polynomial fitting proved to be effective in representing time series, with low errors and high accuracy. These methods were useful for planning and forecasting in the telecommunications sector, as exemplified by the analysis of ARCOTEL users.