Skip to main content

Table 2 Algorithms and parameters considered in the ML/IA analysis

From: Machine learning approach to support taxonomic species discrimination based on helminth collections data

Algorithms MM + GL + H     
Correct instances (%) Kappa Specificity Sensitivity AUC Accuracy NPV PPV
J48 93.172 0.92 0.962 0.966 0.979 0.964 0.965 0.964
Random Tree 89.257 0.89 0.934 0.951 0.944 0.942 0.949 0.936
REPTree 90.963 0.90 0.959 0.945 0.986 0.952 0.943 0.961
LMT 96.385 0.96 0.981 0.982 0.999 0.981 0.981 0.982
Majority Voting 94.679 0.94 0.974 0.970 0.972 0.964 0.969 0.975
Algorithms MM + H      
Correct instances (%) Kappa Specificity Sensitivity AUC Accuracy NPV PPV
J48 88.253 0.88 0.932 0.941 0.955 0.937 0.939 0.934
Random Tree 86.646 0.86 0.949 0.935 0.93 0.942 0.933 0.950
REPTree 85.943 0.85 0.919 0.911 0.979 0.915 0.908 0.921
LMT 93.975 0.94 0.965 0.972 0.998 0.968 0.971 0.966
Majority Voting 91.867 0.91 0.955 0.959 0.957 0.957 0.958 0.956
Algorithms MM + GL      
Correct instances (%) Kappa Specificity Sensitivity AUC Accuracy NPV PPV
J48 92.570 0.92 0.960 0.961 0.975 0.961 0.960 0.961
Random Tree 91.566 0.91 0.951 0.959 0.956 0.955 0.958 0.953
REPTree 89.056 0.88 0.942 0.941 0.98 0.941 0.939 0.944
LMT 96.686 0.96 0.984 0.982 0.999 0.983 0.981 0.985
Majority Voting 95.683 0.95 0.980 0.975 0.978 0.978 0.975 0.980
Algorithms MM      
Correct instances (%) Kappa Specificity Sensitivity AUC Accuracy NPV PPV
J48 84.538 0.84 0.918 0.912 0.912 0.915 0.909 0.920
Random Tree 82.228 0.81 0.885 0.917 0.917 0.901 0.915 0.888
REPTree 84.337 0.83 0.915 0.912 0.912 0.914 0.909 0.918
LMT 91.867 0.91 0.956 0.958 0.958 0.957 0.957 0.957
Majority Voting 89.056 0.88 0.934 0.949 0.949 0.941 0.947 0.936
  1. Performance of algorithms is reported as specificity, sensitivity and accuracy following [24] and as corrected classified instances, kappa coefficient and AUC, as generated by Weka 3.8.3 software
  2. MM morphological and morphometric data, GL geographical location, H host, AUC area under the receiver-operating characteristic (ROC) curve, NPV negative predictive value, PPV positive predictive value