Multi-criteria appraisal recommendation

Generating the overall assessments of cases from their observations on multiple criteria when large volumes of historical data have been accumulated is a key issue. This study, therefore, developed the framework of multi-criteria appraisal recommendation (MCAR). Five strategies belonging to three categories were designed to recommend the overall appraisals of new cases from their observations on multiple criteria based on relevant historical data. The proposed framework's basic conditions and key issues were presented to widen its application. The framework was then used to generate the diagnostic recommendations for thyroid nodules from their observations based on the historical examination reports of six radiologists. The experimental results indicated that different strategies are appropriate for different radiologists, and no single strategy was found to be the most appropriate for all considered radiologists. The five strategies were compared with four representative machine learning models to highlight their performances and interpretabilities using the historical examination reports of the radiologists.