Previsão de empregos na construção civil em Alagoas utilizando séries temporais

Abstract

The construction industry plays a crucial role in the Brazilian economy, significantly contributing to socioeconomic development and employability. In the state of Alagoas, where the sector is one of the main economic forces, employability has been affected by a series of factors, including changes in economic policy, financial crises, and the COVID-19 pandemic. This study proposes forecasting the balance of jobs in the construction industry in Alagoas using time series forecasting techniques, aiming to assist technical and professional training institutions in aligning their training programs with the sector’s demands and policymakers in developing more effective employment and education strategies. Employment data in the sector were obtained from the RAIS and CAGED databases, government tools that collect information about admissions, layoffs, and formal jobs in Brazil. Various forecasting models, including Holt Winters Exponential Smoothing, ARIMA, SARIMA, TBATS, and Prophet, were applied and compared, with hyperparameter optimization and crossvalidation performed to improve forecast accuracy and evaluate model performance. The results highlight Holt Winters Exponential Smoothing as the most effective approach, achieving an RMSE error of 825.26, a reduction of 43.5% compared to the widely used Prophet method, suggesting that this classic technique remains a robust and effective tool for predicting job demand in the construction sector. The study acknowledges the inherent limitations of the univariate approach, which does not take into account exogenous variables, and suggests that future research explore modeling techniques that consider such variables. In summary, this research provides a reliable and effective method for predicting job demand in the construction industry, offering relevant guidance for construction companies, technical and professional training institutions, and policymakers.