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

Table 8

Confusion matrix on the Milan dataset without activities balancing using threefold in the combination of hierarchical and ANN.

Acts	Slp	Tlt	Dsk	Dnr	Gbr	Kch	Mbr	Lh	Mr	Red	Tv	Mmd	Chr	Emd	Med
Slp	82.4							2.3	9.4	2.6	2.3
Tlt		75.5			7.0		17.5
Dsk			83.3	4.5					3.2				9.0
Dnr				80.8		13.3				1.2	4.7
Gbr		9.5			85.5		5.0
Kch				1.0		86.2	1.3					11.5
Mbr	2.0	13.7			3.2		81.1		3.0
Lh						1.0		95.0	4.0
Mr	6.0						2.4		80.0	3.6	6.4		1.6
Red									3.7	91.3	5.0
Tv						4.1			2.9	10.8	82.2
Mmd						10.0	3.1					74.9		12.0
Chr						5.0			15.0		4.0		76.0
Emd						10.3						22.5		67.2
Med					2.5	4.4	5.3	5.8	1.0			7.1		11.5	60.4

The columns represent the predicted activities, while the rows represent the actual activities. The performance of overlapping activities is highlighted in bold.