Application of a Random Subspace Method for the Classification of Hyperspectral Data Yasser Maghsoudi M.Sc student K.N.Toosi University of Technology, Iran Email: ymaghsoudi@yahoo.com Abbas Alimohammadi Assistant Professor K.N.Toosi University of Technology Email: alimoh_abb@yahoo.com Mohammad J. Valadan Zoej Assistant Professor K.N.Toosi University of Technology Email: valadanzouj@kntu.ac.ir Barat Mojaradi Phd Student K.N.Toosi University of Technology Email: Mojaradi@alborz.kntu.ac.ir The improved spectral resolution of modern hyperspectral sensors with its huge amount of spectral data has greatly extended the scope of traditional remote sensing, it not only allows scientists in environmental and geoscience research communities to obtain much more information about different materials on earth, but also provides many challenges for data analysis tasks. However a very high dimensional input space requires an exponentially large amount of data to adequately and reliably represent the classes in that space. On the other hand with increase in the input dimensionality the hypothesis space grows exponentially which makes the classification performance very unreliable. In this paper, an approach based on the concept of Random Subspace Method has been proposed. By using a random sampling of features from the original feature space different feature subsets are generated. The Different feature subsets are randomly selected and passed on to the maximum likelihood classifiers. The outputs of each maximum likelihood classifier are vectors of probabilities for different classes. Four operators including the Max, Min, Mean and Product are then used to combine the outputs of the individual classifiers. In order to find the most optimal size of the subsets we carry out experiments on different subset size. Practical examination of the approach on the AVIRIS data show that the Mean and Product method offers better performance as compared to other operators.