To examine potential relationships between meteorological variables and forest fires in Serbia, daily temperature, precipitation, relative humidity and wind speed data for 15 meteorological stations across Serbia were used to construct fire indices. The daily values of the Ångström and Nesterov indices were calculated for the period 2000-2017. A high number of forest fires occurred in 2007 and 2012 in Serbia, during a period of extremely high air temperatures in 2007, followed by the longest heat wave and the worst drought in 2012. In order to identify the ideal weather conditions for fire break outs, different combinations of input variables, e.g., meteorological variables (mean temperature, precipitation, relative humidity, maximum temperature, minimum temperature and wind speed), fire danger indices or a combination of both, for the Belgrade area during the period 1986-2017, were tested. It was found that using relative humidity or precipitation as a predictor only generates a satisfactory model for forecasting of number of forest fires.