A Hybrid Feature Extraction Method for Nepali COVID-19-Related Tweets Classification

Table 4

Class-wise performance of proposed hybrid features with nine machine learning algorithms (LR, KNN, NB, DT, RF, ETC, AdaBoost, MLP-NN, and SVM).

Classifiers	Positive			Neutral			Negative
	P	R	F	P	R	F	P	R	F
LR	67.9	79.4	73.2	44.1	16.6	24.1	71.4	74.0	72.7
KNN	65.4	79.0	71.5	45.7	15.1	22.7	69.2	69.9	69.9
NB	66.8	56.9	61.4	24.9	51.6	33.6	71.0	55.4	62.3
DT	62.5	63.6	63.0	26.4	21.9	23.9	61.9	64.4	63.1
RF	64.6	84.7	73.3	70.0	10.7	18.6	72.8	69.9	71.3
ETC	63.5	87.7	73.7	73.8	12.2	20.9	75.9	66.6	71.0
AdaBoost	64.2	79.4	71.0	46.3	11.2	18.1	68.4	69.2	68.8
MLP-NN	69.7	76.2	72.8	40.	25.9	31.7	71.	73.6	72.9
SVM + Linear	67.1	80.7	73.3	50.4	09.0	15.3	70.2	75.1	72.6
SVM + RBF	69.7	83.4	75.9	58.7	17.9	27.4	74.4	76.9	75.6

Note that P, R, and F denote Precision, Recall, and F1-score for three classes (Positive, Neutral, and Negative).