A novel fusion model based on deep information mining for wireless propagation modeling

The wireless propagation model is important for accurate 5 G network deployment. However, the traditional wireless propagation model is faced with the problems of limited application scenarios, unstable prediction results and high marginal cost of improving accuracy. In order to solve these problems, this paper constructs new features from the original data from different angles, and uses the random forest model to select the core features, which are used to train the fusion model based on the linear weighted summation of regression models such as KNN, LightGBM, and Bagging. After training, the final fusion model is obtained, it solves the problems faced by traditional wireless propagation models. The results and analysis show that the fusion model outperforms the traditional wireless propagation models and the single models that constitutes the fusion model in terms of prediction accuracy and stability, and is not limited by scenarios and easy to deploy.