A new approach on prediction of fever disease by using a combination of Dempster Shafer and Naïve bayes

Abstract

Health is an important aspect of human life. Symptom of fever is one of the symptoms that can interfere with human health. The symptoms are common in human, but for handling errors sometimes occur diagnosis that can lead to death. Such errors can occur due to lack of expertise or reluctance of patients to check themselves since the symptom fever is common. By considering the issues, we conduct a study to design an application that can help patients with the fever symptom. The research was conducted by combining two following concepts: expert systems (i.e., Dempster Shafer) and machine learning (i.e., Naïve bayes). By combining the methods, we can obtain a single solution considering knowledge formulated by human experts and extracted from data training. Moreover, the application is implemented by an R package connecting R language with PHP, which is RShinny. A case study was taken by using medical records from the hospital Muhammadiyah, Bandung, West Java, Indonesia. To determine the level of accuracy of the system, we carried out two experimental stages, namely the fitting and testing steps. For the fitting step, we obtained the accuracy of 70.67 percent while 56.25 percent is the accuracy of the testing stage.