Breast Carcinoma Prediction Through Integration of Machine Learning Models

• Published in: IEEE access Vol. 12; pp. 134635 - 134650
• Main Authors: Martinez-Licort, Rosmeri, de la Cruz Leon, Carlos, Agarwal, Deevyankar, Sahelices, Benjamin, de la Torre, Isabel, Pablo Miramontes-Gonzalez, Jose, Amoon, Mohammed
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Analytical models; Breast cancer; Data models; Datasets; Diagnosis; Ensemble learning; Machine learning; majority voting; Performance evaluation; Principal component analysis; Public health; Support vector machines; Training
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3431998

Breast cancer poses a global health challenge, with high incidence and mortality rates. Early detection and precise diagnosis are crucial for patient prognosis. Machine learning (ML) models applied to mammary biopsy image data hold promise for achieving an efficient and accurate breast cancer diagnosis. In this study, we evaluated the performance of several ML algorithms, including Logistic Regression (LR), Random Forest (RF), Naive Bayes (NB) and Support Vector Machine (SVM). We establish evaluation contexts by implementing data standardization and reducing the correlation between variables. Firstly, we select the best-performing parameters for each algorithm by building and evaluating the individual models. Then, we implement a combined model using weighted voting, where the weights of each model are determined based on its performance on the test dataset. The final model is constructed by combining the LR, RF and SVM models. We find that SVM is the best-performance individual model, so it has the highest weight in the final model. The final integrated model achieves an accuracy of 98%, a precision of 97%, a recall of 99%, an F1-score of 98% and an AUC of 0.98. Our weighted voting model compares favourably with the other models analysed. This approach demonstrates its efficiency and transparency in handling structured medical data. It is a prototype that will be refined and expanded to encompass larger real-world datasets.