Supervised learning using ANFIS is proposed to identify classes of minerals potential in a satellite based hyperspectral data set. Spectral data for each pixel in a data set are extracted and processed to obtain a characterization map through a novel method. The characterization map is then clustered using c-means fuzzy clustering to obtain characteristic cluster center data. An ANFIS network with three membership function is trained with the cluster center data as input, and a numerical coding of the mineral map of the area as training output. The network satisfactorily learns to identify different classes of mineral and can also indicate the presence of new minerals for which it has not been trained. Such novel mineral can thus be identified and encoded for learning by the network.