PMC:4289553 / 24358-24977 JSONTXT

{"target":"http://pubannotation.org/docs/sourcedb/PMC/sourceid/4289553","sourcedb":"PMC","sourceid":"4289553","source_url":"https://www.ncbi.nlm.nih.gov/pmc/4289553","text":"Modern modelling techniques such as SVM, NN and RF need far more events per variable to achieve a stable AUC-value than classical modelling techniques such as LR and CART. If very large data sets are available, modern techniques such as RF may potentially achieve an AUC-value that exceeds the AUC-values of modelling techniques such as LR. The improvement over simple LR models may, however, be minor, as was shown in the two empirical examples in this study. This implies that modern modelling techniques should only be considered in medical prediction problems if very large data sets with many events are available.","tracks":[]}