Research On User Behavior Analysis And Sleep Staging Based On Data Mining

With the rapid development of technology and the improvement of national living standards,people are increasingly pursuing intelligent lifestyles and healthy living conditions.They are enthusiastic about using various electronic devices to record daily living behavior and monitor health indicators such as heart rate,blood pressure,and electrocardiogram.Detection devices such as smart homes and smart bracelets have received widespread attention,for example,using smart home devices to record the daily living behavior of users and assess whether their living conditions are healthy;Use smart bracelets and other devices to detect basic vital signs such as electrocardiogram and blood pressure of users.In this era of interconnected technology,various monitoring indicators and behavioral records have been widely generated and paid attention to.Therefore,how to discover potential patterns and meaningful information from these large amounts of surface detection data is a problem worthy of attention and research.This thesis explores and proposes data mining based on user behavior analysis methods and sleep staging methods.Firstly,in response to the problems of simple mining content and inaccurate description of user behavior habits in traditional frequent pattern mining methods,this thesis proposes a behavior pattern mining method based on suffix index and time interval.By introducing the concepts of time attributes and time intervals in this method,frequent patterns of user behavior with time information are discovered,and frequent itemsets are obtained through postfix indexing,reducing the memory space occupied by the algorithm when obtaining frequent patterns,accurately characterizing the habits and patterns of users,and having more practical significance.Secondly,in response to the problems of expensive signal acquisition equipment and high professional requirements for sleep staging in commonly used sleep staging methods,which are difficult to achieve daily use,this thesis proposes an automatic sleep staging method based on heart rate variability.Extract heart rate variability signals from daily electrocardiogram signals,and select machine learning sleep staging methods based on sleep features and automatic feature extraction sleep staging methods based on deep learning for experiments.Both sleep staging methods can achieve automatic sleep staging based on heart rate variability,with low professional requirements and suitable for daily and home use.In this thesis,behavior pattern mining and sleep staging experiments were carried out on the open data set,the feasibility of the proposed method based on suffix index and time interval and the method based on heart rate variability is verified.The experimental results show that the proposed method based on suffix index and time interval can mine frequent patterns with time information while ensuring the performance of the algorithm,the model can accurately represent the life habits and rules of the users,and has more practical significance.The sleep staging method based on heart rate variability proposed in this thesis can achieve automatic sleep staging,in which,the accuracy of machine learning sleep staging method based on sleep features is higher than 87%,and the accuracy of automatic feature extraction sleep staging method based on deep learning is higher than 91%,can meet the needs of daily sleep monitoring and professional requirements are low,suitable for daily and home use.