| Summary: | 碩士 === 國立虎尾科技大學 === 工業管理系工業工程與管理碩士班 === 106 === According to the World Health Organization statistics in 2017, more than 15 million people worldwide have suffered strokes, 87% of which are ischemic strokes. Data from the department of health and welfare's top 10 causes of death in 2017 showed that cerebrovascular diseases were the fourth top 10 causes of death in Taiwan, and patients who survived stroke had a high probability of disability, resulting in high medical expenses, care burdens and labor losses. Diseases related to urinary calculi include septicemia, urinary tract infection, cardiovascular diseases, stroke, etc. In addition, patients with stone disease aged 20 to 34 years old have a 1.47 times higher probability of developing stroke than ordinary patients.
This study applies particle swarm optimization algorithm, cross entropy algorithm and genetic algorithm logistic regression, combining case-based reasoning system, construct the evaluation system of urinary calculi patients suffering from stroke, respectively, at the same time combined with support vector machine and back-propagation neural network, construct the prediction model. Finally, the difference of performance of the constructed model was analyzed by using Friedman-Test and Paired Sample T-Test.
The study showed that the six prediction models, with average ACC and AURC reaching 93.8 % and 0.88 respectively, were suitable as the prediction models. The six models were found to have significant differences among the models through the Friedman-Test, among which particle swarm optimization algorithm combined with support vector machine and integrated inverted transfer neural network were superior to other prediction models. Among the three evaluation systems of CBR, all the three algorithms were found to have significant differences by Friedman-Test. The average ACC was more than 85%, and the average AURC was more than 0.84. Particle swarm optimization algorithm combined with case-based reasoning system was the best, the average accuracy rate was 90.76%, and the average AURC was 0.8762. The results of this study can be used as a reference for the diagnostic evaluation of related medical institutions, patients and clinical staff.
|
|---|
