Table 1.

Validation result for each household based on the trained model.

Household ID	Abnormal daysa	Total observed days	Lossb	AUCc	Abnormal weekendd
HH0	66	360	0.18	0.88	43
HH1	60	360	0.22	0.76	15
HH2	35	325	0.22	0.63	7
HH3	53	365	0.13	0.78	12
HH4	63	311	0.13	0.8	37
HH5	73	352	0.17	0.52	61
HH6	62	365	0.23	0.8	16
HH7	68	362	0.21	0.72	31
HH8	73	365	0.17	0.76	32
HH9	50	365	0.2	0.83	24
HH10	61	351	0.21	0.81	26
HH11	61	364	0.18	0.8	23
HH12	99	355	0.19	0.75	68
HH13	65	363	0.21	0.77	38
HH14	70	356	0.19	0.68	21
HH15	61	363	0.19	0.66	23
HH16	74	365	0.2	0.72	39
HH17	79	359	0.22	0.65	20
HH18	60	356	0.18	0.68	15
HH19	49	355	0.18	0.74	19
HH20	71	363	0.21	0.74	45
HH21	61	363	0.2	0.77	13
HH22	86	355	0.16	0.82	50
HH23	79	361	0.21	0.72	37
HH24	46	364	0.28	0.74	18
HH25	65	356	0.15	0.74	17
HH26	92	350	0.17	0.71	55
HH27	70	360	0.18	0.68	29
HH28	61	363	0.21	0.72	29

^aAnomalous activity of users in 2019: activity that deviated from regular activity patterns defined from the 2018 data.

^bAverage error associated with reconstructing the validation records using the regular activity pattern.

^cAUC: area under the curve. Overall compatibility of the regular activity pattern with validation records, in terms of recognizing the activation and deactivation of motion sensors at the right time slots.

^dTotal number of abnormal days that are weekend days.