ORIGINAL PAPER Clustering of mineral prospectivity area as an unsupervised classification approach to explore copper deposit Maysam Abedi & Gholam Hossain Norouzi & Seyed Ali Torabi Received: 18 July 2011 /Accepted: 25 June 2012 # Saudi Society for Geosciences 2012 Abstract This paper describes the usage of clustering meth- ods including self-organizing map (SOM) and fuzzy c- means (FCM) which are applied to prepare mineral prospec- tivity map. Different evidential layers, including geological, geophysical, and geochemical, to evaluate Now Chun cop- per deposit located in the Kerman province of Iran are used. Clustering approaches are used to reduce the dimension of 13 feature vectors derived from different layers. At first, Geospatial Information Systems (GIS) is employed to ana- lyze and integrate different layers, and the area under study is prioritized to five classes. Then, the SOM as an unsuper- vised classification method is carried out to classify this area into five clusters. Produced clusters are compared with GIS prospect map, while the SOM results are matched with the GIS output. The main reason to use the FCM is that a vector belongs simultaneously to more than one cluster so that membership values of each cluster can be mapped. As a consequence, clusters generated by the SOM and FCM are considerably matched with five-class-map of the GIS ap- proach. The chosen cluster as a high potential location to additional drilling is matched to the main alteration and faults zone. To validate generated clusters for mineral po- tential mapping, geological matching of study area and selected proper cluster can be a satisfactory way. Finally, clustering methods can be a very fast approach to interpret the area under study. Keywords Mineral prospectivity mapping . GIS . Self- organizing mapping . Fuzzy c-means clustering . Copper deposit Introduction Mineral exploration is a sophisticated process whose main purpose is to discover new mineral deposits in the region of interest. To achieve this goal, one of the main steps is to demarcate prospective areas. Four this purpose, various thematic (e.g., geological, geophysical, geochemical) geo- data sets should be collected, analyzed, and integrated for mineral prospectivity mapping (MPM). The MPM process is a multi-criteria decision-making task on different scales. Several approaches may be used for MPM, which can be divided into either data-driven or knowledge-driven meth- ods (Bonham-Carter 1994; Pan and Harris 2000; Carranza 2008). In data-driven techniques, the known mineral depos- its are used as “training points” for establishing spatial relationships with particular geological, geochemical, and geophysical features. The spatial relationships between the input data and the training points are quantified and used to establish the importance of each evidence map and finally integrated into a single mineral prospectivity map (Nykänen and Salmirinne 2007; Carranza 2009). Examples of the empirical methods used are weights of evidence (Bonham- Carter et al. 1989), logistic regression (Agterberg and Bonham-Carter 1999), neural networks (Singer and Kouda 1996; Porwal et al. 2003, 2004), and evidential belief func- tions (Carranza and Hale 2002; Carranza 2008). The other techniques, in which a geoscientist’ s expert opinion is applied, are called knowledge-driven methods. M. Abedi (*) : G. H. Norouzi Department of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran e-mail: MaysamAbedi@ut.ac.ir G. H. Norouzi e-mail: norouzih@ut.ac.ir S. A. Torabi Department of Industrial Engineering, College of Engineering, University of Tehran, Tehran, Iran e-mail: satorabi@ut.ac.ir Arab J Geosci DOI 10.1007/s12517-012-0615-5