Research On 5G Cell User Classification Method Based On Improved SVM

With the rapid development of mobile communication technology,5G technology has been widely used.However,the 5G cellular network technology still needs some help in practical applications.The ultra-dense cellular networks increase the amount of physical layer information and the freedom of signal transmission,leading to more severe inter-cell interference.To adopt effective interference coordination technology,the key lies in the classification of central users and edge users.This thesis conducts research on the classification of cell users,as follows efficiently:By analyzing the characteristics of wireless communication systems,a comprehensive classification scheme based on an improved support vector machine(SVM)is proposed to address the shortcomings of the cell user classification scheme with a single parameter empirical threshold.Firstly,some of the easily classified users are classified according to the distance threshold.Secondly,the remaining users are classified twice using the improved SVM algorithm.The study showed that the scheme combines simplicity and classification accuracy.An improved algorithm based on the particle swarm optimization support vector machine(PSOSVM)is proposed to address the shortcomings of the PSO-SVM.Firstly,the training sample data is preprocessed to select the candidate set of support vectors.Secondly,the vector candidates are used as training samples to train the least squares support vector machines(LSSVM),and the optimal parameters are found by the PSO algorithm to obtain the best trained classification model.Finally,the simulation results show that the improved algorithm has higher classification accuracy and shorter training time,providing a solid theoretical basis for classifying cell users.To address the problem of the slow training speed of truncated pinball loss support vector machine(TPin-SVM),an improved algorithm based on TPin-SVM is proposed.Firstly,the TPin-SVM-based optimization model is constructed.Secondly,the solution of the quadratic optimization problem is transformed into the solution of a linear system of equations by combining the idea of least squares.Finally,simulation experiments verify that the improved algorithm improves the training speed without losing the classification prediction accuracy,performs well in the problem of cell user classification,and has good generalization ability.