Table 4.

Summary of the data mining methods used in the reviewed studies.

Data mining method and techniques or algorithms			Data size		Independent variable		Dependent variable
Correlation analysis
	Pearson correlation [20]	Preliminary study: 24 users over 20 days; final study: 19 users over 21 days		TSTa and contextual factors		SOLb, NAWKc, and sleep rating
	Spearman correlation [16]	12 users over 2 weeks		Contextual factors		TST, WASOd, NAWK, SOL, and SEe
	Repeated measure correlation [19]	10 users over 2 weeks		Bedtime, TIBf, and contextual factors		SE, SOL, NAWK, restlessness, TIB, and LSEQg
Regression analysis
	Piecewise fixed effects regression [34]	31,793 users over 18 months; all American users		Time of day, time after waking up, and sleep duration		Cognitive performance
	Simple linear regression [43]	Approximately 20,000 users over 4 months		Contextual factors		SOL, NAWK, and SE
	Linear mixed effects model [35]	557 users over 1 year		Bedtime regularity		Resting heart rate
Rule induction
	A priori algorithm [33]	1 user over 180 days; 4 users over 2 weeks		Contextual factors		SE
	Learn from Examples using Rough Sets [44]	280 users over 1 month; only the data of males were used		Contextual factors		Sleep ratio
	Event mining (+causal inference) [42]	1 user over 800 days		Contextual factors		SOL, WASO, NAWK, and SE
Causal inference
	Stratified propensity score analysis [43]	Approximately 20,000 users over 4 months		Contextual factors		SOL, NAWK, and SE
	Bayesian network analysis [36]	5200 users over 6 months		Contextual factors and bedtime		Contextual factors and bedtime
Time series analysis
	Anomaly detection [37]	1 user over 35 days		Fitbit measured intraday time series, TST, WASO, NAWK, and bedtime		Permutation entropy of sleep time series
	SAXh-based motif matching and principle optimization [38]	100 users over 10 weeks		Heart rate time series data		PSQIi
Statistical test
	Unpaired 2-samples Wilcoxon test [40]	271 users over 8 months		Contextual factors		Statistical differences between good and poor sleep
Decision tree
	J4.8 Classifier [31]	400 users over 15 months		Contextual factors		PSQI

^aTST: total sleep time.

^bSOL: sleep onset latency.

^cNAWK: number of awakenings.

^dWASO: wake after sleep onset.

^eSE: sleep efficiency.

^fTIB: time in bed.

^gLSEQ: Leeds Sleep Evaluation Questionnaire.

^hSAX: Symbolic Aggregate Approximation.

ⁱPSQI: Pittsburgh Sleep Quality Index.