In this study, we evaluate the classification accuracy of inverse synthetic aperture radar (ISAR) images reconstructed using the conventional Fourier transform (FT) and sparse recovery algorithms based on compressive sensing (CS) from incomplete radar cross section (RCS) data. When data are missing from the received RCS dataset, we cannot obtain correct ISAR images using the FT-based method. To alleviate this problem, we propose the use of sparse recovery algorithms. Results show that performing ISAR classification using sparse recovery algorithms can provide reliable classification accuracy, even though the received RCS datasets are incomplete, whereas the FT-based method is unable to do so.