Base Station Traffic Prediction based on Feature Selection and Stacking Ensemble Learning

Accurately predicting base station network traffic is of great significance to improve network service quality and reduce base station operating costs. Aiming at the problem of low prediction accuracy of single model in the existing base station traffic prediction methods, a multi-model fusion prediction method based on feature selection and stacking ensemble learning is proposed. Firstly, a large number of features are constructed on the historical data, and then feature selection and correlation verification are carried out based on the tree model, and the features with high correlation are retained as the input of the predictive model to improve the performance and interpretability of the model. On this basis, a stacking ensemble learning prediction model with GDBT, XGBoost, LightGBM as the base learner and MLP as the meta-learner is established, and finally experimental verification is carried out on the real 1731 base stations. The results show that the mean squared error (MSE) and mean absolute error (MAE) of this method are reduced by 9.8% and 4.3%, respectively, compared with the single machine learning prediction model, and have better prediction accuracy and generalization ability.