Identification of Effective Factors in Breast Cancer Survival in Isfahan Using Machine Learning Techniques

Abstract

Background & Objective: Breast cancer is a leading cause of female mortalities worldwide. This study has used machine learning techniques to determine the most critical factors influencing the survival rate of breast cancer patients in Isfahan. Materials & Methods: A list of variables influencing the survival of breast cancer patients was initially extracted from the data sets of two Isfahan hospitals for this analytical investigation, leading to the extraction of 16 critical factors based on the opinions of oncologists. In the next step, the missing values were identified and deleted or corrected, followed by converting some features into numerical ranges. Ultimately, the key variables influencing the survival rate of breast cancer patients were determined by applying 11 machine learning algorithms. Results: Forward selection is more accurate than other techniques. Of the 15 input features, 13 were extracted as influential survival rates at least once using different techniques, with BC-ER-PR-HER2 ranking first among the features. The six first features, including Bc-ER-PR-HER2, lymph node dissection, behavior, primary surgery procedure, the exact number of nodes examined, and the exact number of positive nodes, were determined as the best combination for identifying breast cancer patients. Even though cancer behavior patterns differ in various societies, there are still similarities in risk factors. Conclusion: Forward selection combined with principal component analysis using support vector machines, neural networks, and random forests can be the best model for breast cancer prediction. Neural networks, random forests, and support vector machines are very good at predicting breast cancer survival. © 2023, J. Adv. Med. Biomed. Res. All rights reserved.