Table 6. Dataset 4 (Date vs. Party hubs) predictions from classifiers trained using machine learning methods.

Approach	Best k	Accuracy	F1 Score	Precision	Recall	C.C.	AUC
NB k-gram	3	67.1	59.8	.54	.67	.33	.71
NB(k)	3	65.1	58.0	.53	.64	.29	.65
Decision Tree	1	53.5	55.4	.50	.62	.08	.53
SVM	3	62.1	59.0	.59	.59	.24	.66
ANN	2	66.2	55.5	.70	.46	.30	.69
Naive Bayes	1	65.2	57.5	.66	.51	.29	.70
Domain-based	N/A	59.1	30.2	.62	.20	.14	.57
Homology-based	N/A	29.8	22.0	.22	.22	−.43	N/A
HybSVM	N/A	69.2	62.6	.71	.56	.37	.68

Accuracy, F-measure (F1 Score), precision, recall, correlation coefficient (C.C.), and area under the receiver operating characteristic curve (AUC) of classification for the multi-interface versus singlish-interface dataset are presented. Accuracy and F-measure are reported in percentage. For each machine learning approach, values of k ranged from 1 to 4. Only the classifier with the best performing k-value (as defined by highest correlation coefficient) is shown. Our methods were estimated by cross-validation. The highest performing value(s) for each performance measure is highlighted in bold.