Effect of feature selection on machine learning algorithms for more accurate predictor of surgical outcomes in Benign Pro Static Hyperplasia cases (BPH)

Predicting the clinical outcome prior to minimally invasive treatments for Benign Prostatic Hperlasia (BPH) cases would be very useful. However, clinical prediction has not been reliable in spite of multiple assessment parameters, such as symptom indices and flow rates. In our prior study, Artificial Intelligence (AI) algorithms were used to train computers to predict the surgical outcome in BPH patients treated by TURP or VLAP. Our aim was to investigate whether, based on eleven clinical biomarker features, AI can reproduce the clinical outcome of known cases and assist the urologist in predicting surgical outcomes. In this paper, the objective is to perform data analysis to investigate if specific features have a greater impact on predicting whether the patients had the desired outcome after a surgical procedure is done. Finally, how the number of significant features ought to be weighted to predict the outcome after surgery, is determined to create the most accurate prediction method. Here both the Decision Tree and Naïve Bayse machine learning methods are used and compared.