ライフエンジニアリング部門シンポジウム2023にて修士学生2名が発表を行いました

9月16日に東洋大学川越キャンパスを会場に開催されたSICE（計測自動制御学会）ライフエンジニアリング部門シンポジウム2023にて、院生2名がそれぞれの研究内容を発表しました。

発表内容

⾝体活動向上のためのパーソナライズ型Just-in-Time介⼊

池ケ⾕舞（発表者）、杉⼭和輝、村⽥⼤河、⼩嶋健汰、⾦鎭赫

The health benefits of physical activity (PA) are well-known, but adherence to regular PA programs remains a major challenge. Justin-time adaptive intervention (JITAI) has been proposed as a novel intervention design to promote PA by adapting to an individual’s lifestyle and physical capacity. However, most studies implementing JITAI have not tailored intervention criteria to the individual. This study examined the effectiveness of JITAI using Personalized vs. Uniform (across all participants) intervention criteria (PIC vs. UIC) for promoting PA. Twenty-eight young adults wore a wrist activity monitor for two weeks. Participants were divided into two groups which received JITAI to promote PA according to either PIC or UIC. In the first week, mean distance moved and sedentary time per hour for each participant were calculated to derive PIC. PIC were averaged over all participants to obtain UIC. In the second week, JITAI prompts were sent every hour if either distance moved was shorter or sedentary time was longer than PIC/UIC. Multilevel modeling showed that both criteria increased PA during the first hour after JITAI, but more increased PA was found with PIC-based JITAI. Use of JITAI (both criteria) did not significantly increase PA above levels in the first week. JITAI based on tailored intervention criteria can transiently increase PA levels. Studies are needed to develop effective long-term intervention designs with sustained effects.

⽇常⽣活下における睡眠タイミングの推定

村⽥⼤河（発表者）、⼭本祐輔、⾦鎭赫

Sleep regularity is known to affect both physical and mental health. In recent years, research using machine learning has been conducted in the field of sleep science, but only a few studies have focused on sleep timing estimation. In this study, we used a wrist wearable device to estimate sleep timing in daily life. The study was conducted over a two-week period on 37 college students (20.7±1.7 '' 13 female) with no health problems. A total of 294 features including sleep, physical activity, and mood were used to estimate bedtime, bed out time, and mid-sleep time. The dataset consisted of 261 days, and these were divided into training and test datasets at a ratio of 8:2. We were able to estimate bedtime, bed out time, and mid-sleep time with an accuracy of 0.56, 0.55, and 0.60, respectively, by random forest analysis. Features related to sleep regularity such as mid-sleep time were selected as the most important features, but it was also confirmed that features related to daily moods and physical activity contribute to estimate sleep timing. The results are expected to be applied to sleep interventions to maintain and improve sleep regularity.