<p>Breast Cancer is emerged as the second major cause of death among women. The death rate can be reduced by detecting the cancer at an early stage. Early detection as well as classification of breast cancer makes the treatment very effective. Many deep learning methods are implemented in medical imaging to enhance the accuracy of classification. At the same time, misclassification of breast lesions results in a high false positive rate which diminishes the overall performance. Hence, this article proposes a novel classification method to detect breast cancer by using a Residual radial kernel support vector-based stain bower search algorithm. The main objective of the proposed algorithm is to assist radiologists in identifying anomalies using medical imagining interpretation. The experimental investigations are done on two different dataset via four different phases namely data pre-processing, data augmentation, as well and classification phases. The stain bower bird algorithm is modified with an initial search strategy to overcome inherent search limitations. The stain bower search algorithm is used to tune the parameters and provide optimal hyperparameter values. The proposed system associates the Residual network, radial basis function, and support vector machinewith the stain bower search method to form the proposed algorithm. The accuracy, sensitivity, and specificity of the proposed model are 98.5%, 97%, and 95.2% respectively which are better when compared to other conventional approaches.</p>