[Retracted] Improving Recognition of Overlapping Activities with Less Interclass Variations in Smart Homes through Clustering-Based Classification

Table 12

Performance evaluation metrics on the Milan dataset with activities balancing using leave-one-day-out cross-validation.

Dataset	Cross-validation	Clustering method	Classification method	Precision (%)	Recall (%)	F score [0, 1]	Accuracy (%)
Milan	Leave one day out	Fuzzy C-means [35]	ANN [45]	93.40	93.20	0.93	93.30
			ET-KNN [47]	89.30	89.40	0.89	89.40
			KNN [48]	87.10	87.20	0.87	87.10
			SMO [46]	83.30	83.80	0.83	83.80
		Hierarchical [42]	ANN	90.50	90.40	0.90	90.60
			ET-KNN	88.10	88.20	0.88	88.40
			KNN	86.20	86.50	0.86	86.70
			SMO	83.10	83.70	0.83	83.80
		K-mean [36]	ANN	87.60	87.10	0.87	87.20
			ET-KNN	84.30	85.20	0.85	85.50
			KNN	81.10	81.30	0.81	81.10
			SMO	78.02	78.20	0.78	78.30
		DBSCAN [37]	ANN	85.70	85.10	0.85	85.10
			ET-KNN	81.20	81.30	0.81	81.30
			KNN	79.10	79.10	0.79	79.50
			SMO	78.20	78.50	0.78	78.20

The precision, recall, and accuracy are in percentages (%), while the range of F score is between [0-1] with 1 being the highest. The highest values are in bold.