This study proposes a new weighting scheme of fuzzy time series (FTS) based on a classical approach. In this method, the weights are assigned to the fuzzy logic relationships (FLRs) based on the classical definition of odds. The forecasting accuracy of the proposed weighting scheme is determined using mean square error (MSE) as the statistical error measure. The odd weighting scheme is applied to simulated data  from a Generalized Auto Regressive Conditional Heteroscedasticity GARCH process and to two real data sets; exchange rate data for Malaysian Ringgit (MYR) to the US Dollar (USD) and daily closing prices of Uganda Securities Exchange (USE). Results obtained show that the new weighting scheme is efficient and outperforms the existing methods in as far as forecasting accuracy is concerned.

weights, fuzzy time series, forecasting, odds, volatility, GARCH.