Antibiotic resistance is one of the biggest threats to global health. According to WHO antibiotic resistance is already at high levels of danger. Causes of antibiotic resistance are errors in diagnosis and treatment. Essential facts of disease diagnosis in primary health care units in Indonesia are conventional methods often used because of a shortage of skilled medical personnel, Conventional diagnosis systems depend on the experience of medical personnel without using complex clinical data. This paper use a clinical decision support system (CDSS) based on medical record data and classification techniques to reduce misdiagnosis. This paper focuses on the diagnosis of typhoid fever disease (TFD) because the amount of antibiotic resistance is very high and is an endemic disease in Indonesia that is difficult to diagnose. This system uses three methods of supervision classification, Naïve Bayes (NB), k-nearest neighbor (kNN), Support Vector Machine (SVM). Then the system was compared and evaluated. The results of the three methods showed that kNN provided 88.7% accuracy, better than other classifications for typhoid fever disease.