Skip to main content
. 2017 May 18;8(29):47816–47830. doi: 10.18632/oncotarget.18001

Table 2. The classification accuracy/AUC of 25 WEKA classifiers using combined multi-parametric histogram and texture attributes in LGG and HGG as well as grade II, III and IV gliomas classification.

25 WEKA classifiers (accuracy/AUC) LGG vs. HGG Grade II, III and IV
Original (28 vs. 92) Smote (100 vs. 100) Original (25 vs. 29 vs. 63) Smote (68 vs. 68 vs. 68)
Bayes BayesNet 0.717/0.743 0.750/0.826 0.667/0.836 0.770/0.880
NaiveBayes 0.742/0.717 0.845/0.874 0.641/0.778 0.750/0.885
Lazy IBk## 0.750/0.638 0.905/0.905 0.718/0.795 0.961##/0.971
LWL 0.733/0.769 0.800/0.833 0.735/0.858 0.642/0.892
Functions LibSVM (linear)** # 0.792/0.690 0.945*/0.945 0.786#/0.838 0.956/0.957
Logistic 0.708/0.698 0.885/0.934 0.556/0.716 0.828/0.895
SimpleLogistic 0.792/0.841 0.930/0.957 0.735/0.888 0.922/0.976
SGD 0.792/0.715 0.930/0.930 0.701/0.816 0.917/0.977
SMO**# 0.758/0.668 0.945*/0.945 0.786#/0.874 0.956/0.975
VotedPerceptron 0.758/0.697 0.800/0.861 0.590/0.758 0.657/0.873
Meta AdaBoostM1* 0.808*/0.793 0.875/0.956 0.675/0.925 0.809/0.894
Bagging 0.783/0.818 0.855/0.933 0.726/0.912 0.858/0.966
ClassificationViaRegression 0.708/0.800 0.830/0.900 0.658/0.879 0.939/0.843
LogitBoost* 0.808*/0.846 0.885/0.945 0.675/0.891 0.877/0.974
Rules Decision Table 0.642/0.597 0.795/0.896 0.761/0.912 0.745/0.871
Jrip 0.767/0.612 0.850/0.808 0.726/0.838 0.814/0.879
OneR 0.792/0.616 0.645/0.645 0.718/0.809 0.672/0.754
PART 0.633/0.567 0.830/0.804 0.692/0.821 0.775/0.818
Trees DecisionStump 0.767/0.629 0.815/0.630 0.726/0.771 0.304/0.608
HoeffdingTree 0.742/0.218 0.850/0.875 0.650/0.777 0.750/0.885
J48 0.675/0.397 0.855/0.801 0.684/0.817 0.833/0.872
LMT 0.800/0.849 0.930/0.958 0.744/0.896 0.922/0.976
RandomForest 0.792/0.845 0.915/0.976 0.752/0.892 0.922/0.984
RandomTree 0.658/0.540 0.815/0.813 0.607/0.698 0.755/0.818
REPTree 0.742/0.460 0.820/0.850 0.650/0.837 0.779/0.901

*, ** represent the classifier with the highest classifying accuracy on original and SMOTE LGG and HGG glioma data, respectively. #, ## represent the classifier with the highest classifying accuracy on original and SMOTE grade II-III-IV glioma data, respectively.