Predicting Sleep Efficiency Based on Interpretable AI Using Smart Wearable Device (Preprint)

• DOI: 10.2196/preprints.70678
• Publication Date: 2025-01-02T22:35:33Z
• Authors: Ji-Hoon Jeong; Minji Lee; Yun Kwan Kim; Yerim Huh; Se-Na Jang; Jun-Young Oh; Young-Jin Kee; Glenn van der Lande; Olivia Gosseries
• Keywords: Preprint; Smartwatch; Wearable Technology; Sleep

<sec> <title>BACKGROUND</title> Sleep plays a crucial role in the immune system, memory, and emotional control, which turn affects overall quality of life. Although several studies have used wearable devices to measure sleep efficiency, these tools primarily monitor patterns without offering strategies improve efficiency. </sec> <title>OBJECTIVE</title> In response, our study introduced framework that not only measures efficiency using technology but also suggests lifestyle changes enhance quality, explainable artificial intelligence. <title>METHODS</title> Lifelog data was measured smartwatch features can be controlled by user, such as number steps, were extracted from lifelog data. A light gradient-boosted machine utilized classifier predict whether low or high. Using Shapley additive explanation, important related walking speed identified lifestyles recommended We investigated relationship between stages, excluded uncontrolled factors devices. <title>RESULTS</title> Results show that, following 10-fold leave-one-subject-out cross-validation evaluation, an accuracy rate 80% achieved measuring scenario where adjusted according feature importance derived model output, improvement bad good observed. addition, stage variables found statistically higher for cases exhibiting <title>CONCLUSIONS</title> This provides insights into developing technologies suggest potentially leading better health outcomes through improved quality.