Abstarct: Deep learning, a subset of artificial intelligence, aims to extract specific features by processing data. This approach has been a novel method for solving theoretical and experimental physics problems in recent years. In this thesis, we use this method to study gauge-gravity duality and examine how the gravitational side can be constructed for a given quantum field theory. To construct gravity, the metric must be determined, meaning that the gravitational metric should be derived using experimental data. In this thesis, we first review the studies conducted to obtain the gravitational metric using deep learning. Then, we calculate the real potential for quarkonium. By analyzing the real components of this potential, we investigate the quarkonium decay process. The findings show that the distance between the quark-antiquark pair, known as the dissociation length, decreases with increasing temperature. After that, we calculate the imaginary part of the potential. The results indicate that the extracted metric can reflect the predictions of quantum chromodynamics theory.