Expression of Concern for: Breast Cancer Screening Using Machine Learning Models

• Published in: 2022 3rd International Conference on Intelligent Engineering and Management (ICIEM) p. 1
• Main Authors: Singh, Gurinder, Chaturvedi, Prateek, Shrivastava, Anurag, Vikram Singh, S.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27-04-2022
• DOI: 10.1109/ICIEM54221.2022.10703467

The main objective of this paper is the development of an Algorithm for Boolean Breast Cancer Classification based on 116 patients with results by testing blood along with physical information (life, BMI, Gluco-level, diabetes, etc.). The following machine learning models were compared for performance in this study: choice tree, arbitrary forest, K-closest neighbors, ANN, AVM, and replace analysis. With the following methodologies, the information was gathered: Data is separated among 80 percent training and 20 percent testing using k-fold cross-validation (k 10). Initially, the average of accuracy, as well as sensitivity, are examined, while the counts of accuracy, sensitivity, specificity, also region subjected to a number of tests were analyzed in the second. Later on, mass measures of mammography are likely to be conducted on terrible cancers. As a result, other techniques could be used in such evaluations to enhance choice, also ML could give a lot of value and an excellent cost-effective ratio in the breast cancer diagnosis strategy. In previous times, various research publications on breast-cancer indicators are been published. The applied strategy would look at two analytical parameters: the linear calibration curve and maturity. The q-parameter, Pearson-correlation analysis, also intake parameters dependency, the experiment only along the parameters having a criterion of the importance of 6% is also calculated with the evaluation of the standard deviation (evaluated with the KS experiment), these could be as Gluco-level, diabetic, resisting, and prototype for assessing equilibrium. The arbitrary forest is utilized for the learning procedure along with 10 factors as the best final classifier, with 84 percent accuracy, 100%_sensitivity, 65%_specificity, also 0.90 curved surface region.