Effect of the Different Membership Function Fitting Methods in Personalized Risk Calculation

Abstract

In patient monitoring, patient-specific evaluation is essential to obtain realistic results. For this reason, the effect of personal characteristics should be incorporated into the system. To ensure this requirement, the number of the input factors, the input factors themselves and their limits should be varied depending on the personal profile. To handle the inputs with no sharp boundaries, fuzzy based inference should be used in the system. In this paper, besides these solutions, previous statistics are also considered during the membership function determination. The aim is to find the most realistic, but also simplest membership function-shape to decrease the computational needs, during the evaluation, while it takes into account the usual reactions of the patient.