Shear strength parameters are important for assessing the stability of structures, and are costly to calculate using conventional methods. In this research, simple geotechnical techniques and artificial intelligence were used to calculate the angle of internal friction and soil cohesion without the need for more complex testing. To this end, intact samples from 14 boreholes in Bandar Abbas, which had undergone primary geotechnical testing and direct cutting, were selected and used to train neural networks. 195 networks were trained in in this research. To achieve the best performance, feedforward neural networks were first trained in single and double layer modes with a low number of neurons in the middle layer, and the TRAIN BR function was selected due to the high ratio of R (0.97). Then, by incorporating additional layers, the Median model was trained using configurations of 3, 4, and 5 layers, each with varying numbers of neurons in the intermediate layer (50, 40, 30, 20, and 10). The results show that the four-layer MLP network gives the best results, for this mode R training 1, the test R is 0.90 and the total R is 0.98. Finally, to validate the neural network, 15 samples were selected and the input parameters of the network were trained in the optimal states of 2, 3, and 4 layers, then the output of the network was evaluated. For cohesion prediction, the neural network in 4-layer mode (R2=0.99) and 2, 3 and 4-layer networks (R2=0.99) have the best output for the friction angle.
