Breast cancer diagnosis using Stochastic Self-Organizing Map and Enlarge C4.5

• Published in: Multimedia tools and applications 2023-05, Vol.82 (12), p.18059-18076
• Main Authors: Jaiswal, Arvind, Kumar, Rajeev
• Format: Article
• Language: English
• Subjects: Algorithms; Breast cancer; Clustering; Computer Communication Networks; Computer Science; Confusion; Data Structures and Information Theory; Multimedia Information Systems; Performance measurement; Predictions; Self organizing maps; Special Purpose and Application-Based Systems
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-14265-1

Timely and accurate Breast Cancer (BC) prediction allows healthcare providers and doctors to take suitable decisions to treat the patients. Thus, this study employed a strategy based on Deep Learning (DL) to diagnose BC. The study intends to cluster the BC data by the proposed Stochastic Self-Organizing Map (SOM) as it has the ability to process complex data. On the other hand, Enlarge C4.5 (E-C4.5) algorithm is introduced to predict the BC cases based on the clustered outcomes. The BC dataset is loaded and pre-processing is performed where dimensionality reduction is executed to select only the relevant features for clustering. This process eases the clustering. Then, clustering is undertaken by the proposed Stochastic Self-Organizing Map. Here, all the identical data are grouped as clusters which make it easy for prediction. Followed by this, Breast Cancer is predicted by the proposed Enlarge-C4.5 algorithm. After this, the predicted results are analysed by comparative analysis through the four standard performance metrics. This analysis is significant as it shows the degree to which the introduced techniques are effective than the existing techniques. The histogram, correlation map, confusion matrix and clustering results are also discussed clearly. The analytical outcomes explore that the proposed methods are effective than the conventional methods as the proposed method shows a high accuracy rate, precision rate, recall rate and F1-score rate. The misinterpretation rate is also found to be minimum on implementing the proposed method, which is confirmed through the confusion matrix. The proposed method also determines the malignant and benign counts.