Deep Learning- and Word Embedding-Based Heterogeneous Classifier Ensembles for Text Classification

Table 5

The list of classification accuracies of single classifiers and heterogeneous ensemble systems with majority voting and stacking integration strategies on datasets represented with TF-IDF and the average of Word2vec vectors.

Methods	20News-18828	20News-19997	Mini-news	Web KB4	1150haber	Milliyet	Hurriyet	Aahaber	Computation time
MVNB	75.83	63.78	77.00	80.21	87.97	83.70	78.80	83.71	1 h 57 m
MNB	90.27	81.74	81.98	87.80	96.78	84.42	83.09	84.80	2 h 24 m
SVM	91.51	64.08	92.35	92.05	94.20	91.36	83.34	81.56	2 h 55 m
RF	91.25	79.15	91.47	90.92	97.99	92.10	85.92	90.38	4 h 5 m
CNN	91.88	77.80	92.12	93.01	95.59	79.49	89.27	94.15	5 h 48 m
Heter-MV	92.90	79.90	92.68	93.75	96.00	92.13	88.53	93.10	8 h 32 m
Heter-Stck	94.30	82.23	95.06	95.08	97.40	93.90	90.02	94.86	9 h 20 m