Biomedical Text Categorization Based on Ensemble Pruning and Optimized Topic Modelling

Table 12

Two-way ANOVA test results of the macro-averaged F-measure.

Statistical analysis of results on different LDA-based configurations
Source	DF	SS	MS	F	P
Configuration	23	0.42777	0.01860	91.27	P<0.001
Dataset	4	5.99867	1.49967	7359.42	P<0.001
Classifier	1	0.09263	0.09263	454.58	P<0.001
ConfigurationDataset	92	0.03536	0.00038	1.89	P<0.001
ConfigurationClassifier	23	0.09800	0.00426	20.91	P<0.001
DatasetClassifier	4	0.01123	0.00281	13.78	P<0.001
Error	92	0.01875	0.00020
Total	239	6.68241
Statistical analysis of results on classifiers and ensemble pruning methods (with LDA (k=50) based representation).
Source	DF	SS	MS	F	P
Configuration	17	0.27733	0.016314	23.26	P<0.001
Dataset	4	2.41143	0.692858	859.46	P<0.001
Error	68	0.04770	0.000701
Total	89	2.73646
Statistical analysis of results on conventional classifiers, ensemble learners, and ensemble pruning methods (with BA-LDA (DB) based representation).
Source	DF	SS	MS	F	P
Configuration	13	0.03613	0.002780	14.68	P<0.001
Dataset	4	1.53718	0.384296	2029.89	P<0.001
Error	52	0.00984	0.000189
Total	69	1.58316