PREDICTION OF OVARIAN CANCER USING ENSEMBLE LEARNING METHODS

ABSTRACT

Ovarian cancer is one of the world's most deadly diseases among women. However, life expectancy of the patients can be prolonged if it is early detected. Although a few techniques, e.g., ultrasonography and clinical IT scanning, have been utilised to distinguish a benign and malignant non-gynaecologic conditions, the tumor biomarkers detection is not an easy one. The main aim of our study is to apply machine learning models along with statistical methods to the clinical data obtained to conduct predictive analytics for the early diagnose. In statistical analysis, Chi-Squared, information gain, and gain ratio are used to filter the best features in finding the significant blood biomarkers. Thereafter a set of machine learning models including naive bayes, k-nearest neighbor, multilayer perceptron, support vector machine, decision tree and logistic regression are used to build classification models to differentiate between benign and malignant ovarian cancer patients. To improve the classification accuracy, the diversity of the model was further modeled using ensemble strategies which includes bagging, boosting and stacking approaches. The results from the predictive analysis suggest that the machine learning models can classify malignant patients from benign patients with high accuracy as good as 96%. This project shows that the ensemble learning models perform better than the classical/traditional single models and since early detection appears to be a challenge, and also improvement of our proposal can be made.