Abstarct: drought is a natural creep and slight phenomenon of any area's climate which entails strong economical ecological and social effects. Iran is a country of dryness and has witnessed numerous detriments from drought condition in recent years. For effective controling and predicting the intensity of drough with high precision and calculating the lack of certainty tool development can help politicians to lessen vulnerabilty and plan percisly to confort this phenomenon. Through last decades, a large number of abilities in series modeling and prediction were shown by neural networks. by this research, drought modeling and prediction will be proceeded by using ANFIS, ANN. by evaluating the uncertainty of each model through next step, the appropriate model will be chosen. as long as input data play a significant role in neural networks instruction and through all recent researches up to now absolutely coincidental selection algorithm was used tried to present a new algorithm in order to choose input data in instruction level. Final results show that although ANFIS model shows less R2 than ANN model but it has less lack of certainty than ANN model which it proves the superioverty of ANFIS model than ANN model.