BREAST CANCER PREDICTION USING SUPPORT VECTOR MACHINE AND RANDOM FOREST ALGORITHMS

Abstract

Breast cancer continues to be one of the primary causes of cancer-related death globally, highlighting the necessity of precise and prompt diagnostic techniques. Despite being widely utilized, conventional methods like mammography, ultrasonography, and biopsy are frequently linked to subjectivity, high false-positive rates, and restricted accessibility in places with limited resources. This study aims to evaluate the effectiveness of Support Vector Machine (SVM) and Random Forest Classifiers (RFC) for breast cancer prediction, with a focus on enhancing diagnostic accuracy through advanced feature extraction and dimensionality reduction. The Kaggle repository provided a publicly accessible dataset with 569 occurrences and 33 characteristics. Principal Component Analysis (PCA) was used to minimize dimensionality while maintaining 90–95% of the variance in the data. Standard performance indicators, such as accuracy, precision, recall, F1-score, and ROC AUC, were used to train and assess the model. Confusion matrix analysis, ROC curves, and classification reports provided additional support. According to the results, SVM performed the best, with an accuracy of 97.7% and a ROC AUC of 0.99. RFC came in second with an accuracy of 94.2% and a ROC AUC of 0.98. Both SVM and RFC demonstrated a good balance between recall and precision, reducing diagnostic error.

DOI: https://doi.org/10.5281/zenodo.21473365