The carbonate reservoirs in the Tarim Basin are characterized by anisotropy and strong heterogeneity. Combined with an integrated analysis of data from seismic, geology, and drilling results, a series of attributes which are suitable for fractured and caved carbonate reservoir prediction is discussed, including amplitude, coherence analysis, spectra decomposition, seismic absorption attenuation analysis and impedance inversion. Moreover, 3-D optimization of these attributes is achieved by integration of multivariate discriminant analysis and principle component analysis, where the logging data are taken as training samples. Using the optimized results, the spatial distribution and configuration features of the caved reservoirs can be chara cterized in detail. This technique not only improves the understanding of the spatial distribution of current reservoirs but also provides a signifi cant basis for the discovery and production of carbonate reservoirs in the Tarim Basin.