Machine Learning Assisted Decision Support System for Prediction of Prostrate Cancer

Abstract

Over the past several years, there has been a global rise in the prevalence of prostate cancer. It was discovered that prostate cancer is the most often diagnosed cancer category amongst men and it can be stated as the main cause of cancer-related mortality worldwide among males. Diagnosing illnesses is one of the greatest obstacles in medicine. This study was crucial due to the lack of precise standards for the evaluation of prostate cancer symptoms and the low predictive accuracy of current diagnostic approaches. It is believed that machine learning approaches may be used to solve situations when there are no precise and defined rules and where the event-influencing aspects can be predicted. Computer-aided systems produce a variety of solutions with this knowledge. In this study, the performance of various supervised machine learning algorithms (SVC, LR, AdaBoost (Ada B), XG Boost (XGB), KNC, LGBM, GB, DT, and RF) is compared and discussed. In this study, we acquired data from Kaggle consisting of 100 cases and 10 characteristics. In our model, we initially determined the maximum accuracy for XGB, LGBM, and RF to be 93.33 percent. Eventually, we used GridsearchCV to tune hyperparameters in order to improve the performance of the classifiers. This time, the highest accuracy was determined to be 96.67% not just for those three, but also for GB as a whole. The most noteworthy finding of this study is the improvement in accuracy and consistency of predictions. Therefore, if the computer is educated with machine learning methods using patient data, it can be therapeutically beneficial in predicting cancer with a high degree of accuracy. In this method, an unnecessary patient biopsy can be avoided.