Research On User Controlling Behavior Mining Algorithm With Time Factor For Smart Home

With the advent of the era of big data,the smart home industry has been greatly developed.However,the large amount of user historical controlling data has not been well utilized,which limits the competitiveness of the industry.Therefore,it is one of the key issues in the process of smart home industry moving towards intelligent to learn how to use the vast amount of smart home user historical controlling data to mine the user controlling behaviors and realize the intelligent decision of home equipment.According to the standards in the smart home industry,this paper divides the user controlling behavior into two categories: user controlling behavior for single device and user controlling behavior for multiple devices.Aiming at the current mainstream user controlling behavior mining algorithm cannot well utilize the time characteristics of user historical controlling data,a user controlling behavior mining algorithm for single device and a user controlling behavior mining algorithm for multiple devices have been proposed,and the verification experiment was performed.The main research work of this paper is as follows:(1)Introduced the concept of smart home and user controlling behavior,and focused on the clustering analysis and association analysis techniques currently used for smart home user controlling behavior mining.(2)Summarized and analyzed the clustering analysis algorithms which are often used for mining user controlling behavior for single device,and proposed a self-organizing user controlling behavior mining algorithm with forgetting learning ability for single smart home device.Firstly,aiming at the problem that the current clustering analysis algorithms lack the self-organizing ability to select the clusters number and the self-organizing clustering process is too slow,this algorithm proposes an artificial neural network that combines self-encoding technology and competitive learning mechanism to initialize the algorithm to generate a more reasonable clusters number and corresponding centroid.Secondly,concerned with the problem that the current cluster analysis algorithms cannot integrate the time factor to judge the importance of user historical controlling data and lackthe ability to do forgetting learning,this algorithm introduces a forgetting factor based on Ebbinghaus forgetting law to improve the centroid update mechanism of the traditional cluster analysis,so that the algorithm can perform the ability to do brain-forgetting learning and mine the user controlling behaviors that are more familiar to the current situation.(3)Summarized and analyzed the association analysis algorithms which are often used for mining user controlling behavior for multiple devices,and a user association controlling behavior mining algorithm with time factor has been proposed.The current association analysis algorithms only use the support degree and the confidence degree as the screening conditions in the process of mining user association controlling behavior and fail to fully benefit the time characteristics of the user historical controlling data,thus the operations included in the user association controlling behavior lack of order and relevance in the time dimension.In view of this,this algorithm first obtains the time-series frequent itemsets of user historical controlling data through the improved FP-growth algorithm and then introduces a time constraint factor to constrain the process of generating user association controlling behavior from time-series frequent itemsets.It ensures that there are a sequence and relevance in the time dimension between the operations included in the user association controlling behavior.(4)Introduced how to apply the proposed user controlling behavior mining algorithms to the smart home system.Based on the cloud smart home system,a smart home control software and a user controlling behavior mining software were developed.Firstly,remote monitoring of smart home devices and collection of user historical controlling data are realized through the control software.Then,the user controlling behavior mining software will be used to mine and recommend the user controlling behaviors.Finally,the simulation experiment shows that the proposed user controlling behavior mining algorithms have feasibility and engineering application value.