Risk prediction platform for pancreatic fistula after pancreatoduodenectomy using artificial intelligence

Table 2 Prediction performance of the various dataset for postoperative pancreatic fistula

Dataset and algorithm	Number of original variables	Number of encoded variables	Number of samples	Area under curve
Random forest with complete cases	38	44	889	0.670.02
Neural network with complete cases				0.740.02
Random forest with complete variables	34	40	1769	0.670.01
Neural network with complete variables				0.720.02
Random forest with missing data treatment	38	44	1769	0.680.02
Neural network with missing data treatment				0.710.02