Research On Base Station Traffic Prediction And Network Planning Algorithms Based On Big Data

With the prosperity of big data technology and new network technology,wireless network has undergone tremendous changes,showing the characteristics of intensiveness,complexity,flexibility and dynamics,and provides more and more diversified types of services.The required quality of service is gradually increasing.Therefore,how to analyze and plan complex wireless networks becomes an important research task.However,traditional network planning methods have poor performance in terms of accuracy and efficiency,which cannot meet the high-quality requirements of various new network plans.Therefore,combining with the analysis method of big data,this thesis deeply studies the intelligent wireless network planning methods.Based on the introduction of the existing network planning methods,this thesis mainly carries out the following three aspects:(1)A base station traffic prediction model based on space-time analysis is proposed.From the perspective of time and space,taking base station coverage as unit area,this method learns the relationship between base station human flow distribution and the human-flow transfer between the base stations,and comprehensively analyses the spatial and temporal characteristics of base station human flow distribution and the interaction characteristics of human flow between base stations,so as to propose a prediction method for base station human flow.The simulation results show that this method makes up for the shortcomings of traditional prediction methods,improves the accuracy of traffic prediction,and provides effective guidance for later network planning.(2)A network planning method based on data mining is proposed for RPMA(Random Phase Multiple Access)low-power wide-area network with large density of base stations and uneven traffic distribution.First,a signal quality prediction model is established by using the boosting regression trees algorithm,which is used to extract the coverage distribution spacial pattern of the network.Then,the weighted K-centroids clustering algorithm is utilized to obtain the optimal base station deployment for the current spacial pattern.Finally,according to the total objective function,the best base station topology is determined.Experiment results with the real data sets show that compared with the traditional network planning method,the proposed method can improve the coverage of low-power wide-area networks.(3)A demonstration platform of network planning system based on big data is designed and implemented.The platform is built on the utility of Jupter Notebook.It mainly uses Python libraries such as scikit-learning and pandas to realize the preprocessing of network data,the training and prediction of signal quality model,andthe clustering adjustment process of base station locations.In this part,the construction of the whole platform and the implementation of the core planning algorithm are introduced in detail.Based on this,the effectiveness of the proposed network planning method is verified.