Abstract:
Innovation in the health sector in the world is advancing, and there are increasingly more challenges that need to be
addressed. The current problem is that most of the world's population is afflicted with diseases, such as heart disease, that go
undiagnosed because of their subtle symptoms and the difficulty and cost of diagnostic techniques. Many struggle to access adequate
medical treatment and expensive diagnostic testing. The only function of medical devices is to monitor health data, and no diagnostic
procedure aids patients. As an application of system diagnosis in IoT, this research develops a diagnosis system, provides treatment
information, and an appointment system in health data processing. This study uses machine learning and fuzzy logic approaches to
offer convenience to patients in self-diagnosis processes monitored by doctors. To optimize the IoT product, a fuzzy logic
experiment was tested to produce a diagnosis with three variable parameters: stress level, oxygen, and temperature. These three
variables will diagnose disease symptoms experienced by users based on the measurement of four data sensors: heart rate, oxygen
saturation, galvanic skin response, and body temperature. In the machine learning approach, the experiment conducted trials with
several Decision Tree, KNN, SVM, Random Forest, and Logistic Regression models to forecast cardiovascular disease diagnosis.
The Confusion Matrix results show that the approach with the highest value is Random Forest, with an Precision of 81.5%, Recall of
83.7%, F1-Score 82.5%, and Accuracy of 82.6%. This indicates that diagnosing heart disease can be more efficient using the
Random Forest approach. With these two approaches, patients can be facilitated in carrying out the diagnosis process independently
and remotely without the need to come to the hospital. Doctors can easily monitor and provide treatment with each patient's
electronic health record platforms. This is expected to increase the level of optimal health services.
