O uso do google trends para melhorar a previsão de séries temporais: o caso da taxa de desocupação no Brasil

Abstract

Among the applications of Time Series Analysis, forecasting future values is one of the most used. In forecasts, the accuracy of the method used is one of the critical factors in its adoption. This article aims to verify whether a dynamic regression using Google Trends improves the performance of forecasts for the monthly unemployment rate series of people aged 14 years and over in Brazil, relative to exponential smoothing and ARIMA methods, measured by the mean absolute error and the root mean squared error. The conclusion was that the use of Google Trends index with the chosen keyword showed no gains in its adoption in terms of improving the accuracy of forecasts for the series in question.