A Spatio-Temporal Traffic Forecasting Model for Base Station in Cellular Network

Data-based service becomes the new mainstream of the telecommunication market, since the commercial using of 4G. Thus, one of the greatest challenges is to handle the huge amount of traffic generated by these services. And an accurate forecast of traffic could help with the load balance, resource allocation and network optimization. In this paper, we proposed a spatio-temporal traffic forecasting model to forecast traffic of base station in cellular network. To obtain the temporal and spatial model, we adopted a clustering algorithm based on artificial neural network to build individual models for different types of base stations. Also, we designed a spatial model to deal with the irregular distribution of base stations. Besides, we classified the base stations according to the cluster results of temporal and spatial models, and constructed a linear combination of these two models’ forecasting results for each type of base station. Finally, we evaluated our model on the dataset of a real city in China. The results show that our proposed model makes a good performance than other existing studies.