Research On User Behavior Recognition And Prediction Algorithm Based On Multi-source Data For Smart Home

According to the demands in the smart home industry,this article analyses user behavior with two types of smart home data sources: the first category is user behavior recognition for smart phone sensors;the second category is user behavior prediction for smart device sensors.In view of the current smart home user behavior research algorithm failed to well design the learning model according to the characteristics of the data set,a corresponding smart home user behavior recognition and behavior prediction algorithm have been proposed,and the experimental results were analyzed in detail.The main research work of this paper is as follows:(1)The article thesis introduced the concept and the data collection method of smart home separately.It focused on the data processing technology and behavior learning model currently used in smart home sensors,and analyzed the characteristics of smart phone sensor data and smart device sensor data.(2)This article summarized and analyzed the shortcomings of the current behavior recognition algorithm that applied to smartphone sensors,then proposed a multi-channel one-dimensional convolution residual network behavior recognition algorithm.First of all,in order to solve the problem of feature missing in current behavior recognition methods,the algorithm extracts data features with multiple single spatial dimension convolution kernel based on 3-dimensional features of smartphone sensor data.Secondly,for the degradation of the network model as the number of layers increases,the algorithm adopts a residual network structure to learn user behavior characteristics,so as to classify user behavior activities.Finally,experiments are conducted on two public data sets,attaching with the experimental results of the algorithm in detail with class activation maps.(3)Aiming on the home user behavior prediction methods that currently exists the problem of low accuracy,poor versatility and for lacking of humanization sense,this article proposed a smart home user behavior prediction method based on Bi GRU-DAtt model,which owns two characteristics of power law distribution and symmetry according to the smart home user manipulation behavior data.Firstly,this method adopts the bidirectional Gated Recurrent Unit neural network to figure out the relationship between user manipulation behaviors.Then,it focus on the symmetry control behavior within a certain range by using the attention mechanism.Finally,it compares these experiments with the real user control data in order to show the facts that this method can fully mining the association between user-controlled intelligent devices and user’s behavior habits,and achieves high accuracy user behavior prediction.