Shape Memory Alloys (SMAs), as a branch of smart material actuators, are widely researched in the areas of control applications. These actuators exhibit considerable hysteresis between the supply voltage (conventionally used in resistive heating) and position characteristics of the SMA. Unless a model matches the actuators nonlinearities, the control of an SMA would result in an error between the desired and actual strain. An Adaptive Neuro-Fuzzy Inference System (or ANFIS) model is proposed to model the hysteresis of the system. The hysteresis of an SMA is path dependent, thus controlling the SMA in real-time requires a time series forecasting a nonlinear model. The input parameters for such an ANFIS model would be a physical variable at time t and at a time t-n, where n is a time delay. The present work studies the effect of time delay on the actuator nonlinearities for two ANFIS models. One of the models studies the relationship between the desired displacement of an SMA and the supply voltage across the SMA, while the other model predicts the actual displacement of an SMA from the feedback temperature.