Research On Optimization And Application Of Extreme Learning Machine

Machine learning plays an important role in industrial modeling process, especially in the nonlinear time-varying systems. Among numerous machine learning algorithms, Extreme Learning Machine(ELM) has been widely used in the field of function fitting and classification, for its advantages of fast training speed and good generalization performance, but in the process of dealing with unbalanced data and batch data, Extreme Learning Machine is far from perfect precision. In order to solve the above problems, Weighted Extreme Learning Machine(WELM) and Online Sequential Extreme Learning Machine (OS-ELM) came into being. In the WELM algorithm, the weight is an empirical value, in addition, the input weight and bias of WELM are randomly given, if we use the traditional WELM and OS-ELM processing data, the accuracy is low. With the complexity of the electric power communication network, the amount of power communication data is growing exponentially, which forms the large data of power communication, and the power communication network is a nonlinear time-varying system. To ensure the reliability of this system, we need to find a better performance algorithm. Based on the above backgrounds, the main work of this paper is as follows:(1) In order to optimize the weights, the input weights and the bias of Weighted Extreme Learning Machine, the particle swarm optimization algorithm is applied to the Weighted Extreme Learning Machine, and Weighted Extreme Learning Machine based on PSO(PSO-WELM) is proposed. In this paper, we use the data from the UCI database to do the experiment, the results show that the accuracy of this algorithm is improved.(2) In order to improve the precision of Online Sequential Extreme Learning Machine algorithm, this paper proposed an Online Sequential Extreme Learning Machine based on PSO(PSO-OS-ELM), Using the data from UCI database to do the experiment, the results show that this method has improved the accuracy of the traditional online sequential extreme learning machine.(3) Electric power communication network is a nonlinear time-varying network, general machine learning algorithms can not achieve good learning performance. Based on the advantages of this algorithm, we use the method to identify the key points and the bandwidth of the power communication network, and compared with other algorithms. It is proved that the algorithm is effective in solving the problem of power communication network.