The Prediction of Breast Biopsy Outcomes Using Two Data Mining Algorithms Based on Parameter Variations

Objective: Breast cancer is the type of cancer that causes the most death in women in the United States after lung cancer. The objective of this study was to predict breast biopsy results using age, BI-RADS, Mass Shape, Mass Margin, Mass Density by Multilayer Perceptron and Random Forest algorithms. Material and Methods: The dataset contains a BI-RADS assessment, the patient’s age and three BI-RADS attributes together with the ground truth for 516 benign and 445 malignant masses. WEKA software was used for data mining. The data mining methods of the Multilayer Perceptron and Random Forest were used to predict the severity of cancer. Results: Accuracy, F-measure and Root Mean Squared Error values of the Multilayer Perceptron model were found 0.837, 0.833 and 0.352, respectively while accuracy, F-measure and Root Mean Squared Error values of the Random Forest model were found 0.816, 0.814 and 0.396, respectively. The Multilayer Perceptron method provided a better prediction of breast cancer diagnosis than the Random Forest method and a software was developed based on the best model created by the Multilayer Perceptron method. Conclusion: Consequently, a model that was built with the MLP method can be used as an alternative in the diagnosis of patients and be an assistant tool for physicians.