Applying Machine Learning to Design and Evaluate the White Noise Recommendation System for Insomniacs

Abstract

Many sleep problems have occurred due to changes in the modern lifestyle Insomnia is relatively more serious than other sleep disorders An insomnia not handled properly will increase the risk of depression obesity and cardiovascular diseases Nowadays most sleep therapies involve treatment with drugs which cause side effects on some patients For non-drug therapies sound is used to assist sleep and its research has a solid foundation However sound is quite subjective and it is difficult to determine the sound suitable for each individual Therefore it cannot be effectively applied to insomnia With the diversified development of artificial intelligence experiments with insomnia are rarely conducted through actual products or services Besides as mobile phones' application is quite extensive more discussion is needed for user experience Therefore based on the above this research designs a white noise streaming recommendation system application It uses project-based collaborative filtering to recommend appropriate white noise to patients with insomnia to improve their sleep quality In the end we successfully built the system prototype Then we used a randomized controlled method for 16 patients with mild insomnia to conduct a five-day sleep assessment experiment It is used to verify the effectiveness of the system We use nonparametric tests for statistics The experimental group's average deep sleep and the control group were 25 4% and 21 7% respectively while the average of the REM period was 27 4% and 23 4% This result indicates that it can help improve the stable deep sleep and REM period of insomniacs At the same time we also used System Usability Scale (SUS) and semi-structured interviews to evaluate the usability of 8 people with insomnia In order to understand the usability and willingness of the system This result shows that the system's usability score is 85 which means a good usability state The final research results show that our proposed machine learning recommended white noise solution can help patients with insomnia find suitable white noise in a good experience And this program can help improve their sleep quality This research aims to design a white noise recommendation system that can improve insomnia patients and help improve insomnia problems Finally this research can reference for related research in the future such as sound and sleep human-computer interaction and machine learning

Date of Award	2021
Original language	English
Supervisor	Yu-Hsiu Hung (Supervisor)