Sleep Quality Assessment Analysis Research Based On Machine Learning

Sleep is essential for people's health and well-being.For people of all ages,good sleep contributes to good health and a rich spiritual world.Although the key to people's physical health is regular sleeping time and adequate sleep every day,many people still experience sleep disorders and insomnia.Therefore,there are many wearable devices and smart phones on the market that can monitor sleep status at any time.In particular,smart phones are increasingly becoming a health device because they contain many sensors and people carry them with them every day.Therefore,smart phones are very suitable for long-term sleep monitoring and assessment of individual sleep quality.This article first randomly recruited 30 participants from university students to collect one month of experimental data.However,due to the improper daily operation of some participants,some of the data were abnormal.At this time,the data was preprocessed.,Delete abnormal data,and finally retained the data of 12 participants for experimental analysis.At the same time,a sleep quality questionnaire was produced based on the Pittsburgh Sleep Quality Index as a benchmark for experimental results.This article then features the collected data and extracts a total of 13 features.In order to assess the quality of people's sleep(that is,to distinguish between people who are generally good or poor sleepers),the experiment first performs sleep monitoring and then performs a sleep quality assessment.According to the 13 features of the structure,feature selection was performed,and the optimal features capable of sleep monitoring were selected.At this time,the experiment compares the classification performance of naive Bayes and C4.5 decision trees.The result is that the classification performance of the C4.5 decision tree is relatively high,so the C4.5decision tree is used for feature selection and sleep monitoring.In the feature selection of this paper,experiments were conducted to compare the performance of single feature classification,the comparison of the classification performance of two feature combinations,the comparison of the classificationperformance of three feature combinations,and the comparison of the classification performance of four feature combinations.The best combination of features in different situations.Finally,using the C4.5 decision tree to compare the classification performance of the above four characteristics obtained by the experiment in the best combination of the case,the experiment obtained in the use of the standard deviation of the acceleration change,the average value of the light intensity and the standard deviation of the screen proximity three When the features are combined,the classification performance is optimal.Therefore,these three characteristics are selected for sleep monitoring in the experiment.The sample data is divided into sleep and waking state.At the same time,the sleep time,wake-up time and sleep duration are obtained and the sleep time and Error Analysis of Sleep Duration Prediction.Finally,according to the selected sleep time,wake-up time and sleep duration,the supervised learning model and unsupervised learning model in machine learning were used to evaluate sleep quality.From the experimental results,it can be concluded that the use of the random forest model in the supervised learning model is better for the evaluation of sleep quality.