LAUSR

Abstract The paper offers a crossbreed streamlining algorithm combining harmony search (HS) and simulated annealing (SA) known as harmony search and simulated annealing (HS-SA) for precise and accurate breast malignancy. Additionally, an improved wavelet-based contourlet transform (WBCT) system for feature extraction explores to get the highlights of the region of interest (ROI), permitting performance improvement over other standard methodologies. In the mined feature space, the projected HS-SA algorithm intends to diminish the feature dimensions and congregate at the unprecedented feature set. The SVM classifier backed with diverse kernel functions is used for classification, which is fed by the chosen features, and its exhibition contrasts with the conventional machine learning classification and optimization techniques. The actualized computer-aided diagnosis (CAD) learning mechanism is challenged by evaluating its findings. It examines two different breast mammographic datasets i) benchmark BCDR-F03 dataset and ii) local mammographic dataset. Trial reproductions, empirical outcomes, and measurable examinations likewise indicate that the proposed model is practical and advantageous for the arrangement of malignant breast growth. The findings show that the proposed CAD framework (HS-SA + kernel SVM) is better than different characterization accuracy procedures (with an accuracy of 99.89% for the local mammographic dataset and 99.76% for benchmark BCDR-F03 dataset, AUC of 99.41% for the local mammographic dataset and 99.21% for reference BCDR-F03 dataset), while keeping the feature space limited to just seven feature subsets and computational prerequisites as low as is prudent.