Research On The Method Of Smart Home Pattern Recognition Based On Data Warehouse

With the rapid development of information technology, the Internet of things technology making the direct communication between the machine and machine or between machine and people possible, gradually become a hot research topic. The smart home industry deriving from the Internet of things is also accelerating. The smart home applying information technology to people’s daily life, provides some related services to the daily life, such as security, entertainment, food, and improves the convenience of everyday life. In the past few years a lot of corresponding products have appeared in the market, but most of them only realize the function of integration control. Thus, there is much room to promote intelligent degree of the smart home.In order to promote the intelligence degree of smart home system, this paper designed a method by building a system collecting data produced in users’daily life and introducing double layers ART1Neural Network Algorithm (NNA) for data processing. This system preprocesses collected data to get significant data sets. Double layers ART1NNA algorithm helps the system extract users’interests and preferences in large data sets and transform it into a mature and stable pattern, which reflecting different users’ life model. Thus, with this system, smart home system can be intelligent and closed-loop ecosystem.Users produce huge amount of data daily. In order to help the smart home system manage user data better, this paper adopts a distributed system based on Hadoop, which is a open system for the data storage and efficient framework for data processing, including HDFS, Hive, Map Reduce. With the help of HDFS system, massive user behavior data can be stored stably. Combining Hive with the double ART1NNA method implements the effective user behavior pattern classification. Through matching with the behavior pattern, the user’s later behavior can be predicted. In the lab, the proposed method is verified and the predicted results conform to the user behavior characteristics.