摘要：Patent analysis data has been reason for studies to be an important indicator of technology, i.e. because patents can serve as quantitative indicators to innovation and collaboration. Knowing the evolution of a country of patent applications, can identify how this country is concerned about the technological development. This work aims to apply appropriate techniques to describe the behavior of the number of patent applications for inventions and utility models at the National Institute for Industrial Property (INPI) of Brazil and find relevant periodicity in the data. The numbers of patent applications for inventions and utility models were analyzed from 1996 to 2013. To model the series we used the methodology of Holt-Winters (series without trend, linear trend, exponential trend and models without seasonality, seasonality additive and seasonality multiplicative). The model validation criterion was the MAPE (Mean Absolute Percentage Error). After application of all Holt-Winters models, it has been found that the best model for the model was analyzed series linear trend and multiplicative seasonality, whose MAPE was 6.63% with the parameters ?=0.4, ?= 0,1 and ?=0,1. The forecasting found has a behavior very similarly to real values. This confirms that the data have been adjusted well.
