Using an innovative method for breast cancer diagnosis based on Extreme Gradient Boost optimized by Simplified Memory Bounded A

• Published in: Biomedical signal processing and control Vol. 87; p. 105450
• Main Authors: Chen, Tao, Zhou, Xichao, Wang, Guan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.01.2024
• Subjects: Breast cancer detection; Deep learning; Extreme Gradient Boost; Machine learning; Simplified Memory Bounded A; Deep learning; Breast cancer detection; Extreme Gradient Boost; Simplified Memory Bounded A; Machine learning
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2023.105450

•Presenting an innovative method of BC detection based on ML techniques.•Proposing Extreme Gradient Boost optimized by Simplified Memory Bounded A* for BC detection.•XGBClassifier with 98% recall and precision for BC.•Selecting SGD, GP, Ada Boost, Linear SVC, MLP for diagnosis processes.•Benefits of the computer-aided techniques and supervised learning. Breast cancer (BC) is a prevalent form of cancer among women and is responsible for a significant number of cancer-related deaths. Early detection of BC is crucial in preventing its progression and reducing mortality rates. Accurately classifying tumors as either benign or malignant helps avoid unnecessary treatments. In recent years, computer-aided techniques have gained popularity for pattern recognition and predictive modeling due to their ability to detect important features. This paper introduces a novel machine learning approach for BC diagnosis. Six machine learning methods, including Stochastic Gradient Descent (SGD), Gaussian Process (GP), Ada Boost, Support Vector Classifier (SVC), Multi-Layer Perceptron (MLP), and Extreme Gradient Boost (XGBoost), are employed to classify the dataset. The performance of these methods is evaluated using metrics such as accuracy, precision, recall, and F1-score. To improve classification performance, the dataset is oversampled using the random oversampler technique. Additionally, the selected method is combined with five optimizers to further enhance performance. Among the methods evaluated, XGBoost achieves an accuracy of 96.48% and an F1-score of 95.11% on the base dataset. After applying the oversampling process, these metrics improve to 98.17% and 98.19%, respectively. The inclusion of optimizers enhances the performance of XGBoost, with the XGBoost + SMA mixed method demonstrating the best performance, achieving an accuracy of 98.45% and an F1-score of 98.47%. The results indicate the potential of this method (combination of XGBoost and S-MBA* algorithm) in improving the accuracy and efficiency of breast cancer diagnosis, contributing to early detection and improved patient outcomes.