Smart Home System Design And Implementation Based On Machine Learning

As people increasingly high requirements for the quality of life and the widespread adoption of smart phones, the concept of smart home has been more and more discussed. Integration of electronic devices have become more sophisticated, smart equipment are also emerging, traditional appliances are becoming intelligent, smart home is becoming hot.Smart home is a house that has a variety of sensors and control devices, It can monitor environments and control electrical appliances. It is easier for people to learn more about the family or the environment and control equipment. By that people can get a more comfortable living environment and ensure the safety of their home.There has been a lot of smart home system can provide some of the environmental monitoring and remote control of appliances, but inside the system, the devices are not well bonded together, can only provide separately monitoring and controlling each sensor or node, It is not really intelligent. In recent years, artificial intelligence, machine learning are developing rapidly, so that the smart home has a new chance and opportunity to become useful by using the new technology.The system in this thesis will regularly read all system data and use Tensor Flow which is Google’s newest open source machine learning framework to implement a feed forward neural network, FNN that can analyze the historical data. This thesis use Re Lu as activation function. By continuous operation of the system, we can get the predicted state of each node. When reaching limitation, the system will send control commands, requesting to modify the state of the current node. Combining the smart home and machine learning technology, the whole traditional smart home system can maximize the usage of the system data and provide users with security and convenience for their own habits.By combining data from all nodes and predicting each node, we can get the possible states of each node, by predicting all nodes, the status of all nodes will form a new complete system model. After that, we will make a holistic smart home system, but also can be well maintained flexible, in order to increase or decrease the nodes prepare accordingly.