Optimization of Mamdani-type Fuzzy Risk Assessment Models

Abstract

Nowadays, risk assessment has become the focus of many research efforts, which is the result of environmental, health and/or war related situations. One of the most important trends for the risk assessment models is the use of different patient surveillance systems, which can be used for various purposes, from performance monitoring of athletes to remote monitoring of the elderly. If the aim of the monitoring is to determine the current risk levels, based on the measured values, special care is required. Two fundamental requirements for these kinds of patient monitoring systems are personalization and timely availability of results. However, the proper personalization and the appropriate accuracy, but quickly available result are still difficult issues for researchers. In this article, the authors provide an overview of the methods they propose to address the above questions. In terms of customization, a personal profile-based evaluation is proposed that works with the patient's health characteristics. The applicability of the statistics, generated from previous measurements, in the evaluation is also presented, as well as a method for handling interactions between the input factors. In order to improve the reaction time, the authors propose some methods modifying the traditional Mamdani evaluation, and applying the Higher Order Singular Values Decomposition (HOSVD) method.