Selecting a computational classifier to develop a clinical decision support system (CDSS)

The objective to develop clinical decision support system (CDSS) tools is to help physicians making faster and more reliable clinical decisions. The first step in their development is choose a machine learning classifier as the system core. Previous works reported implementation of artificial neural networks, support vector machines, genetic algorithms, etc. as core classifiers for CDSS; however, these works do not report the parameters considered or the selection process for their implemented classifier. This paper is focus on the selection of a classifier to develop a CDSS. The options were reduced by reviewing advantages and disadvantages of each classifier, comparing them and considering the project parameters. The results of the analysis that take into consideration the project parameters show that some classifiers could be affected negatively in their performance, leaving support vector machines as the more suitable classifier to develop the CDSS by concluding that SVM disadvantages do not affect the accuracy and its advantages are not affected negatively by the project parameters.