Abstarct: Depleting the surface mineral resources has been led to the rising demand for exploring deep deposits. Although there are various mextal mineralization belts in Iran, particularly the Urmia-Dokhtar belt, the explorations have mostly been performed in shallow depths. This issue was due to several reasons, including the detection problems and the exploration of blind economic mineralization. The geochemical zonality method is known as one of the substantial approaches in mining geochemistry. This technique is characterized as one of the most reliable methods in exploring blind economic reserves. However, the shortcoming of this method and the advances in earth data science have forced the researchers to develop integrated approaches by combining novel and traditional techniques to improve the performance and cover the obstacles of the traditional methods. This study proposes a comprehensive methodology to integrate the deep learning technique as an interesting topic in earth data science and the traditional geochemical zonality method. Thus, the shortcomings of the geochemical zonality method are eliminated, and the blind economic mineralization areas are detected more efficiently. The suggested comprehensive methodology deals with seven continuous stages. Every stage plays a significant role in extracting the non-linear and complex geochemical features and the latent relationship among geochemical data. This procedure is conducted by the supervised and unsupervised deep learning algorithms. The performance of the proposed comprehensive methodology has been examined using two challenging areas (i.e., Sungun and Astamal areas) situated in the Ahar-Arasbaran belt. Sungun area is known as a world-class Cu-Mo deposit. Astamal area is a weak area of blind economic mineralization, while it contains strong alterations and Copper mineral indices. The findings demonstrated that the Sungun area had two blind economic mineralization zones in the southeast and northwest. Also, blind mineralization in the Astamal area was recognized in the west and southeast zones. However, these were zone dispersed mineralization and non-economic. This study proved that the proposed comprehensive methodology was an efficient approach to split the geochemical data at different levels of anomalies without computing anomaly matrix and threshold, detecting the blind mineralization areas, determining the type of mineralization (i.e., economic and non-economic), and specifying the drilling location.