Jump to content

Research:Automatic new article topics suggestion/Statistics

From Meta, a Wikimedia project coordination wiki

OneVsRest classification[edit]

Random Forest[edit]

Precision with max-depth 3[edit]

precision.macro params
0.3516 n_estimators=160, max_features="log2", max_depth=3
0.3434 n_estimators=80, max_features="log2", max_depth=3
0.3332 n_estimators=40, max_features="log2", max_depth=3
0.3093 n_estimators=20, max_features="log2", max_depth=3
0.2761 n_estimators=10, max_features="log2", max_depth=3

Precision with n_estimators 160[edit]

precision.macro params
0.3514 max_depth=3, max_features="log2", n_estimators=160
0.3455 max_depth=3, max_features="sqrt", n_estimators=160
0.3256 max_depth=2, max_features="log2", n_estimators=160
0.3163 max_depth=2, max_features="sqrt", n_estimators=160

ROC-AUC with max_depth 3[edit]

roc_auc.macro params
0.9332 max_features="log2", max_depth=3, n_estimators=80
0.9301 max_features="log2", max_depth=3, n_estimators=40
0.9243 max_features="log2", max_depth=3, n_estimators=20
0.913 max_features="log2", max_depth=3, n_estimators=10

ROC-AUC with n_estimators 160[edit]

roc_auc.macro params
0.9404 max_depth=4, max_features="sqrt", n_estimators=160
0.9401 max_depth=4, max_features="log2", n_estimators=160
0.9344 max_depth=3, max_features="log2", n_estimators=160
0.9342 max_depth=3, max_features="sqrt", n_estimators=160
0.9279 max_depth=2, max_features="log2", n_estimators=160
0.9262 max_depth=2, max_features="sqrt", n_estimators=160
0.918 max_depth=1, max_features="log2", n_estimators=160
0.9148 max_depth=1, max_features="sqrt", n_estimators=160

Recall with max_depth 3[edit]

recall.macro params
0.8258 n_estimators=160, max_features="log2", max_depth=3
0.8232 n_estimators=80, max_features="log2", max_depth=3
0.8188 n_estimators=40, max_features="log2", max_depth=3
0.8148 n_estimators=20, max_features="log2", max_depth=3
0.8057 n_estimators=10, max_features="log2", max_depth=3

Recall with n_estimators 100[edit]

recall.macro params
0.8267 max_features="log2", n_estimators=100, max_depth=4
0.8242 max_features="log2", n_estimators=100, max_depth=3
0.8135 max_features="log2", n_estimators=100, max_depth=2
0.795 max_features="log2", n_estimators=100, max_depth=1

Graphs[edit]

Precision/Recall curve[edit]

PR-AUC curve with random forest classifier











ROC-AUC curve with random forest classifier