Due to the fast development of Internet of Everything, there is a rapid rise in the electronic data producing by appliances. For long time data collection, data operation, data analysis and applications will cause big data. To solve these problems, the main purpose of this paper is using RHadoop to build Bayesian regression model. The appliance data are collected from smart meter, and converted into power features. After identifying the state of power data by the state identification method, the system will build regression model. The dependent variable is appliance use time (weeks) and the independent variable is power feature. The score model and evaluate model is to decide which power feature is most suitable for being independent variable at last. The technology is used to explore the correlation between appliance use time and decline curve and to predict appliance use time, in order to enhance the overall behavior analysis in Smart Home.