Analysis Of LTE Network Uplink Interference Based On Improved Machine Learning Algorithm

With the continuous growth of the construction scale and the number of users of the LTE network,the mobile communication network is facing challenges such as capacity,spectrum shortage and so on.In order to meet the increasing demands of users for network quality,adjusting and optimizing the network and improving the efficiency of network utilization have become the main tasks of communication operators after the construction and deployment of mobile communication networks.The continuous development of mobile communication network and the innovation of new services such as mobile Internet and Internet of things are also promoting the reform of network optimization technology.Traditional network optimization technology uses drive test,call quality test,network management center measurement report and user feedback as the main means,drive test and call quality test can accurately solve the problems existing in the network to a certain extent,but the process of data collection consumes a lot of time,manpower and material resources,and the collection cycle is long,so the sudden situation of the network can not be fed back in time.MR measurement report can collect user data in real time,including all user information,can accurately reflect network coverage and operation,and provide data support for network quality evaluation and network optimization,but MR data has a large amount of information,so it is difficult to give full play to data advantages simply by manual analysis.Therefore,this thesis introduces machine learning algorithms into the network optimization work,and uses the powerful learning ability of the algorithm to embed expert experience and knowledge into the model to realize the intelligent optimization of the mobile communication network.First of all,an analysis model combining stack noise reduction stacked denoising autoencoder and extreme learning machine is established based on MR data,the uplink interference power of LTE network is extracted as data set.The stacked denoising autoencoder(SDAE)is trained unsupervisedly to extract high-level abstract features of the data and provide initial parameters for the ELM(Extreme Learning Machine)classifier.This model takes advantage of the fast convergence of ELM and the noise suppression of SDAE,and at the same time overcomes the problem of insufficient robustness caused by random assignment of ELM parameters.Experimental results show that this model improves the efficiency of LTE network uplink interference analysis,and at the same time has strong robustness.Secondly,a multi-classifier combined LTE network uplink interference analysis model with adaptive weights is established.Using the complementarity of Support Vector Machine(SVM)and Random Forest(RF)classifiers in the uplink interference classification of LTE networks,the two classifiers are linearly combined through the adaptive weight method,dynamically adjust the weights of the classifiers according to the eigenvalues of the output probability matrix of each classifier,fully combining the advantages of the two classifiers.Experimental results show that the classification accuracy of the combined classifier for LTE network uplink interference classification is better than that of a single classifier.At the same time,in order to further evaluate the performance of the model in the interference investigation work,this thesis verifies the model recognition results through manual on-site investigation.Through actual network data verification,the LTE network uplink interference analysis model established in this thesis can accurately classify interference,assist staff in network optimization work in rapid and accurate interference investigation,improve the quality and efficiency of optimization work,and ensure the normal operation of the network.