| With the rapid development of science and technology,smart home products based on machine learning and highlighting human behavior activity recognition are constantly understood.While smart home products are basically based on manual operation and automatic control,lacking certain flexibility.In view of this situation,this thesis studies and analyzes smart home dataset based on machine learning methods,uses numpy and pandas libraries for data processing,combines the random forest model in supervised learning algorithms,applies them to the research of smart home system.Through the analysis and visualization of the smart home dataset,the corresponding design basis of the smart home system is obtained.1.In view of the current problems in smart home,such as adversarial machine learning and other network attacks,serious homogeneity of device functions,focusing on single-product functions and ignoring linkage,difficulty in adapting to the common needs of people of different age structures in the family.Therefore,this thesis starts from the aspects of eliminating security risks,refined and integrated design,and responding to user emotions,so as to construct a set of smart home system architecture based on activity recognition and highlight user behavior recognition.In addition,the Zig Bee protocol is used to enhance security and stability,and the random forest model is used for training.2.Analyze different smart home system modules,mainly from the four aspects of environmental perception,intelligent control,safety protection and lighting intelligent control of different tasks in the home to carry out a reasonable division of labor,different subsystems adopt corresponding intelligent technology,so as to achieve overall space and scene control.Then through the processing and analysis of the smart home dataset,the different human behavior data are counted individually and the two features are visualized to obtain the importance of the corresponding features.3.Combined with the analysis results of smart home dataset,starting from the basic spatial layout of smart home.Database analysis,random forest model and correlation analysis are used to design the overall smart home system.Firstly,build a smart home system,user-oriented,form a connection with the intelligent control end,terminal and gateway,realize different functional control through different space partitions.Secondly,the six smart home scenarios are individually designed and connected to multiple scenarios.Finally,the deep control of the user and the system is completed through voice,touch and somatosensory interaction,the connection of different spaces and functions of the mobile terminal is completed by means of interface design,and then evaluate the design.After integrating machine learning methods into the design of smart home system combined with data analysis and design evaluation,the result show good characteristics and it can also improve the design of smart home system. |
