Mammography is the most perceptive method for the detection of early breast cancer. The abnormalities of breast are analyzed by digital mammogram images and the most important indicators of breast malignancy are microcalcifications and masses. An efficient Computer Aided Diagnosis (CAD) system for breast cancer classification is proposed in this study based on Discrete Wavelet Transform (DWT), Symmetric Stochastic Neighbor Embedding (SSNE) and Support Vector Machine (SVM) using digital mammogram images. Two technical approaches are employed for feature selection from the wavelet decomposed mammogram for classification. They are based on the application of SSNE over the decomposed image. At first, SSNE is applied to the whole wavelet decomposed image whereas in the second technique it is applied to individual sub band of the wavelet decomposed image. The whole mammogram classification system is implemented in two consecutive stages. The first stage of the proposed system classifies the mammogram image into normal or abnormal. The severity of the predicted abnormality is further classified either it is benign or malignant associated with mass or microcalcification images. The performance of the proposed mammogram classification system is evaluated using Mammographic Image Analysis Society (MIAS) database images.