In this study, it has been shown that the variation of the real and imaginary components of the complex dielectric function of low density polyethylene (LDPE) with both the polyaniline (PANI) additive and temperature can be predicted with high accuracy by generalized regression neural networks (GRNN) method. First of all, pure LDPE and 0.7 wt.%, 1 wt.% and 3 wt.% PANI doped LDPE/PANI composites were prepared and the variations of epsilon' and epsilon '' components of the samples with frequency were determined at 20 degrees C, 50 degrees C and 80 degrees C by dielectric spectroscopy. Then, with the help of the actual values versus predicted values of the dielectric parameters graphics, the success performance of the GRNN model for determining the related parameters was determined as R-epsilon' = 0.9998 and R-epsilon '' = 0.9365 . From this point, the GRNN model was first used to estimate the variation of the epsilon' and epsilon" components with respect to frequency at 35 degrees C, 65 degrees C and 95 degrees C for the existing samples. Then, the variations of the epsilon' and epsilon '' components with frequency of the two composites (1.5 % and 6 %PANT doped LDPE), which were not produced experimentally, were proposed at 35 degrees C, 50 degrees C, 65 degrees C, 80 degrees C and 95 degrees C by GRNN model. Thus, the dielectric parameters depending on the temperature and frequency for these samples, which have not been produced experimentally, were determined.