DefinePK

Abstract

Previous research shown that men seminal fertility rates significantly decreased in the last twenty years due to health status, life habit and environment. Thus, seminal quality prediction in men fertility has become very demanding decision support systems in the biomedical engineering field. Existing solutions focused on producing the accuracy of the prediction model. They also acknowledged that there were problems of imbalance class, ambiguity and noise of the sample dataset. However, the real issues of the domain were still vague. A framework consists of Resample-Smote Balance with the state-of-art Random Forest (RF) method is proposed to overcome the real-world problem and tested on the Fertility dataset. This method is introduced to alleviate the ambiguity and noise of the sample set. Subsequently, the Smote method is applied to balance the size of the dataset before the classification phase using RF is taken place. The performance of the proposed model is compared with other state-of-art classifiers such as MLP, SVM and DT. Consequently, this work manages to produce the best accuracy model with 98.2% that can improve the ambiguity and noise from the sample set at the same time competent to handle the imbalance class of the dataset. The RF also boost the accuracy model compare with other classifiers due to its capability to produce the most probable class from its majority-voting task as output.