COMPARATIVE ANALYSIS OF SELECTED MACHINE LEARNING MODELS FOR THE PREDICTION OF OVARIAN CANCER

Abstract

Ovarian cancer is among the most common cancers and represents one of the top causes of cancer death in women. One of the main ways to enhance survival of patients with ovarian cancer is to catch it early, but this is difficult since ovarian cancer is asymptomatic in its early stages. Accurate and rapid prediction of ovarian tumors would be useful. Building a model based upon artificial intelligence methods could be one acceptable and accurate method of detecting and predicting this cancer. As previous research has shown, Machine Learning (ML) techniques can support early cancer detection efforts. However, the abilities of these techniques hindered by false positive rates and overfitting, and also by other challenges in generalizing a model. In this study, we will review the performance of six popular algorithms - Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Logistic Regression (LR), Naive Bayes (NB), and K-Nearest Neighbors (KNN) - for the prediction of ovarian cancer with a dataset of 349 patients, which includes 49 clinical and biochemical features from pathology diagnoses. In medicine, the accuracy of interpretation is vital, as misinterpretation could lead to adverse health outcomes. Therefore, all models were examined in terms of accuracy, precision, recall, and F1 score. The Naïve Bayes model performed best overall with the highest accuracy (87.14 %) and precision (87.58 %), whereas the Random Forest model demonstrated the strongest recall (87.56 %). Logistic Regression showed comparable performance across all metrics, whilst K-Nearest Neighbour performed the weakest in all metrics evaluated. These findings support the possibility for ML algorithms, Naïve Bayes in particular, and Random Forest, to be used in improving early detection of ovarian cancer. In conclusion, the study demonstrates the promise of ML for the advancement of the diagnosis of cancer and ultimately improving patient care.