Skip to main content
. 2016 Mar 10;7(3):e152. doi: 10.1038/ctg.2016.11

Table 3. Confusion matrices associated with random forest and logistic classification models.

Random forest model Error ratea Logistic model Error ratea
Colon data set
Primary data set        Secondary data set      
    Predicted       Predicted  
    Normal mucosa Carcinoma       Normal mucosa Carcinoma  
Actual Normal mucosa 384 29   Actual Normal mucosa 83 7  
  Carcinoma 19 561     Carcinoma 6 104  
        4.83%         6.50%
Pairs withheld from model fitting        Pairs withheld from model fitting      
    Predicted       Predicted  
    Normal mucosa Carcinoma       Normal mucosa Carcinoma  
Actual Normal mucosa 389 27   Actual Normal mucosa 73 5  
  Carcinoma 10 372     Carcinoma 3 76  
        4.64%         5.10%
                   
Rectal data set
Primary data set       Secondary data set      
    Predicted       Predicted  
    Normal mucosa Carcinoma       Normal mucosa Carcinoma  
Actual Normal mucosa 213 6   Actual Normal mucosa 87 3  
  Carcinoma 8 333     Carcinoma 4 106  
        2.50%         3.50%
 Pairs withheld from model fitting       Pairs withheld from model fitting      
    Predicted       Predicted  
    Normal mucosa Carcinoma       Normal mucosa Carcinoma  
Actual Normal mucosa 211 11   Actual Normal mucosa 69 2  
  Carcinoma 4 201     Carcinoma 3 84  
        3.51%         3.16%

OOB, out-of-bag.

a

Error rates for primary data sets are OOB estimates; error rates for secondary data sets are leave-one-out estimates.