Beam - column connection is a critical part of frame structure where it has to deal with different types of loads and transmit them. The beam-column connection effects on strength and serviceability of the structure should be taken into consideration throughout the design process. The beam-column connection has major roles in resisting lateral loads like earthquake, wind and blast. Undoubtedly, keeping joints sustain through these loads performing on a structure will protect human lives. For this specific reason, this research was carried out to investigate the beam-column connection by gathering the results from previous experimental researches. These researches have conducted an experimental trial on beam-column joint with different strengthen technique; such as ferrocement and carbon Fiber Reinforced Polymer, or using different types of stirrups like rectangle confining or spiral confining concrete. Theoretical analysis was operated using the finite element software, which is formulated considering the cyclic loading effects. The structural behavior under cyclic loading such as; energy dissipation capacity, stiffness degradation scalar, stress, good self-cantering, good ductility, compressive damage, tensile damage, displacements, equivalent plastic strain and plastic dissipation energy density were demonstrated. Comparisons with experimental results are performed to make sure that the finite element analysis is accurate. The parametric study in the next step will depend on evaluate parameters by calculating errors, accuracy, and predict its behavior by deep learning which considered to be advanced technology procedure of neural networks. In the end, the correlations between these parameters were presented as a prediction equation for parameters, and the best reinforcing details with minimum errors were proposed. For best details reinforcement from unconventional strengthen method was sample DCM- DOUBLE then DCM- SINGLE respectively, both show good handling for Damage dissipation energy density (DMENER), Magnitude of Plastic Strain (PEMAG) & Plastic Dissipation Energy Density (PENER), while for low Scaler Stiffness Degradation (SDEG) value samples (ND-T1 & ND-T2). Deep learning can be used to build equation connect all parameters with minimum error which improved by this research.