Model evaluation dashboard

Breast cancer data, 569 tumors (212 malignant = positive). Stratified 10-fold cross-validation; mean ± standard deviation.

model	accuracy	precision	recall	F1	ROC AUC
logistic regression	0.974 ± 0.018	0.991 ± 0.018	0.939 ± 0.050	0.963 ± 0.026	0.996 ± 0.007
k-NN (k = 7)	0.961 ± 0.027	0.987 ± 0.028	0.911 ± 0.079	0.945 ± 0.041	0.986 ± 0.013
decision tree	0.942 ± 0.035	0.942 ± 0.047	0.902 ± 0.067	0.920 ± 0.047	0.936 ± 0.049
random forest	0.963 ± 0.024	0.959 ± 0.041	0.944 ± 0.041	0.951 ± 0.031	0.991 ± 0.012
SVM (RBF)	0.975 ± 0.014	0.977 ± 0.030	0.958 ± 0.043	0.967 ± 0.019	0.997 ± 0.005
naive Bayes	0.938 ± 0.034	0.944 ± 0.055	0.892 ± 0.075	0.915 ± 0.047	0.987 ± 0.011

logistic regression at threshold 0.5

	predicted malignant	predicted benign
malignant	199	13
benign	2	355

threshold	recall	precision	F1
0.1	0.995	0.784	0.877
0.2	0.981	0.878	0.927
0.3	0.972	0.936	0.954
0.5	0.939	0.990	0.964
0.7	0.887	1.000	0.940
0.9	0.774	1.000	0.872