| With the rapid development of information industry,the Internet of Things(IoT)has become more and more widely used as an important part of the new generation of information technology.The classification and analysis of the large amount of data generated by it is an important part of the IoT technology,which can better exert the application value of the IoT data,and is also a powerful supplement to the IoT.Due to the complexity of IoT applications and some random factors in the generation and transmission of IoT data,which makes IoT data tend to have higher dimensions and more noise,which brings a challenge to the IoT data classification technology.Owing to the unique learning process and strong fitting ability,the Brain Emotional Learning(BEL)model has certain advantages in classification accuracy compared with the traditional classification model.But due to its structural characteristics and the limitations of the training algorithm,it’s poorly performed in high-dimensional data classification and greatly affected by noise.To solve the highdimensional problem,the SA-BEL model is proposed based on Simulated Annealing(SA).By improving the network structure of the BEL model and optimizing the training algorithm of it based on the SA algorithm,the fitting ability and classification accuracy of high-dimensional data will be improved.To solve the multi-noise problem,the Bagging algorithm is used to optimize the SA-BEL model to reduce the proportion of noise samples in training.The model’s anti-noise ability is improved.In the experimental part,several sets of data sets commonly used for algorithm performance testing are selected for comparative analysis.The results show that the SA-BEL model has obvious advantages in classification accuracy compared with the traditional classification model,and it is classification accuracy rate in highdimensional data is improved by 4%~7% compared with the existing BEL model.The SA-BEL model optimized by the Bagging algorithm is less affected by noise,and can maintain the classification accuracy of more than 80% when the noise rate reaches 30%.Finally,the application analysis of the wind IoT data classification problem is carried out to verify the classification ability of the model. |
