Pronóstico de migración humana del Ecuador, utilizando modelos de series temporales

Abstract

The article presents a study to generate forecasts on human migration in Ecuador using time series techniques. The objectives set out were a) To forecast human migration in Ecuador using time series models, b) To conduct a state-of-the-art study and literature review on data mining and machine learning methodologies, c) To evaluate and compare different time series models using precision and performance measures, aiming to select the model that offers the best predictions for future human migration in Ecuador d) To forecast migration in Ecuador using time series models. Initially, the context of migration in Ecuador is presented, which has been affected in recent years by several factors such as earthquakes, changes in government, and the COVID-19 pandemic. The article then delves into theoretical concepts related to time series, components of a time series (trend, seasonality, cycles), and forecasting methods like ARIMA and regression models. The study applies the CRISP-DM methodology using migration data from the National Institute of Statistics and Censuses of Ecuador between 2019-2022. Two models were applied: linear regression and seasonal ARIMA. The results show that the linear regression model had a lower prediction error, so it was chosen as the definitive model. The article concludes that time series models are a valuable tool for making forecasts on social topics such as migration.

Key words: migration, prognosis, CRISP-DM, ARIMA