Application of prediction models using fuzzy sets: A Bayesian inspired approach

A fuzzy inference framework based on fuzzy relations is developed, adapted and applied to temperature and humidity measurements from a specific coffee crop site in Brazil. This framework consists of fuzzy relations over possibility distributions, resulting in a model analogous to a Bayesian inference process. The application of this fuzzy model to a data set of experimental measurements and its correspondent forecasts of temperature and humidity resulted in a set of revised forecasts, that incorporate information from a historical record of the problem. Each set of revised forecasts was compared with the correspondent set of experimental data using two different statistical measures, MAPE (Mean Absolute Percentage Error) and Willmott's D. This comparison showed that the sets of forecasts revised by the fuzzy model exhibited better results than the original forecasts on both statistical measures for more than two thirds of the evaluated cases.