Incoming data prediction in smart home environment with HMM-based machine learning

The Internet of things is characterized by a high level of heterogeneity between its diversified systems ranging from entertainment to automation process. A smart home application is intrinsically dynamic in the sense that it makes up a time series, whose behavior may change over time. The challenge of the incoming data prediction in a smart home is to analyze the energy consumption of each appliance and to notify the risks to remotely control the installed wireless sensor network. This paper proposes a new methodology of data mining in order to predict energy consumption, environment parameters and moving (presence) cases. We present a prediction model based on a hidden-Markov model based for the smart home environment. This model is used as a classification machine learning but it has never used for the incoming data prediction in a smart home. Using a real “Smart Life” database, we demonstrate the validity of our methodology in the scenarios of smart homes incoming data prediction. The proposed prediction technique is tested and proves that there is a high amount of reliability on the considered model.