Breast cancer continues to pose a major global health challenge, with a steady increase in both incidence and mortality projected over the coming decades. While deep learning (DL) technologies have significantly advanced the field of medical diagnostics, most existing approaches depend heavily on labeled datasets, which are often expensive and labor-intensive to produce. In response to this limitation, we introduce DBNALLRBM (Deep Belief Network with All Layers RBM); a framework that performs unsupervised representation learning by stacking Restricted Boltzmann Machines (RBMs) within a Deep Belief Network (DBN) architecture. The model learns hierarchical features directly from unlabeled data, thereby avoiding supervised fine-tuning and manual feature engineering. To assess the discriminative quality of the learned representations, we conducted an evaluation using supervised classification metrics on the Breast Cancer Wisconsin dataset. Our experiments demonstrated strong performance: 97.5% accuracy, 98.46% precision, 94.12% recall, and an F1-score of 96.24%. These results indicate that DBNALLRBM provides a viable and efficient alternative to fully supervised methods, particularly in scenarios where labeled medical data is scarce or costly. More broadly, this study highlights the potential of combining unsupervised generative learning with supervised evaluation to advance diagnostic applications in resource-constrained environments.