As the number of elderly drivers continues to increase, there is a need to identify parameters that can negatively influence their driving performance. Since major parts of fatalities and injuries to elderly drivers occur at intersections, it is evident that improving safety at intersections will decrease the number of dangerous crashes for this age group. This research aims to determine significant parameters associated with drivers’ gap acceptance behavior while they perform a turning maneuver at a four-legged, permissive signalized intersection. For this purpose, drivers of different age group were asked to perform left-turn and right-turn maneuvers at four-legged, permitted signalized intersections developed using driving simulation. Human characteristics of drivers (age, gender, and driving experience), presence of pedestrian in or nearby of the crosswalk, number of lanes, different crosswalk configurations (ladder or standard) and contextual conditions (heavy fog, and night conditions) were considered for generating driving scenarios. The distance between a turning driver (participant’s vehicle) and the nearest on-coming entity (vehicle or pedestrian) was considered as a measurement for how conservative a driver is. A standard linear regression model, Artificial Neural Network (ANN), and the M5’ tree model were employed to identify a correlation between the explanatory variables and the distance to the nearest on-coming entity (as dependent variable). The results illustrated that the age of driver, accepted gap size, and number of lanes; are significantly correlated with the distance to the entity in both left- and right- turn models. Moreover, the results of left-turn models illustrated the importance of other two variables of driver’s gender and presence of pedestrian(s) on the distance to the entity. The results also showed that ANN model outperforms the other two models in producing accurate results; however, this model performs like a black-box and lacks coefficients that are interpretable. This issue of ANN method was addressed by M5’ model. The produced M5’ model not only benefits from the advantages of data mining methods, but it also presents some interpretable formulae to make the model applicable for other data sets.