The exact recognition of breast cancer disease utilizing histology pictures is a difficult assignment, because of the variety of generous tissue and heterogeneity of cell development. In this exploration, a proper component choice and classification methods are proposed for programmed bosom malignancy discovery and characterization. At first, the cores and non-cores cells are portioned from the histological pictures by utilizing Fuzzy C Means (FCM) grouping algorithm. At that point, the component vectors from the sectioned cores and non-cores cells are separated by Speeded up Robust Features (SURF) and shading second highlights. Moreover, Modified Ant Lion Optimizer (MALO) calculation is used to choose the dynamic or ideal element vectors from the removed highlights.

In MALO calculation, competition choice system is utilized to choose the people from the irregular populace to maximize the assembly rate that assists in accomplishing better characterization. At long last, the Capsule Network Architecture (CNA) is used to group the breast cancer disease as benign or malignant. The BreaKHis and Stanford Tissue Microarray Dataset (TMAD) are utilized to research the suggesed model presentation. The division and order execution of the suggested model is assessed by methods for exactness, review, f-score, accuracy, jaccard and dice coefficient. In the trial segment, the suggested model enhanced least of 0.17% and limit of 8.04% of exactness in BreaKHis, and TMAD identified with the current models.