Application Of Deep Learning Method In Intelligent Wireless Propagation Model

With the emergence and development of 5G technology,people have put forward higher requirements on the reference signal receiving power.More and more scholars have joined the study of wireless propagation models in order to better deploy the network,select the appropriate signal sites and divide signal cells.Traditional wireless propagation models can be divided into empirical models,theoretical models,and improved empirical models according to research methods.When models are established,it is often necessary to first divide the propagation scene.Each scene corresponds to a specific model.However,these models often have low accuracy in practical applications.And it is necessary to collect a large amount of field data to correct the model by the least square method.Therefore,building a suitable traditional wireless propagation model is usually time-consuming and labor-intensive.And the accuracy of the model cannot be guaranteed.In order to improve the applicability and accuracy of the model and reduce the cost of model establishment,this paper uses the collected historical data to establish an intelligent wireless propagation model based on deep learning method.This model not only saves the revision cost of traditional wireless propagation models,but also enables faster predictions of reference signal receiving power in specific locations in the environment.The data set used in this paper is Huawei's wireless propagation data set.In combination with the traditional wireless propagation model,14 features are designed.They are the actual height of the transmitter,the horizontal distance between the transmitter and the signal receiver,the actual downtilt of the signal line,signal receiving height,etc.After preprocessing the data,the feature comprehensive index is calculated.This index is based on the divergence of the features,the correlation with the predictors,and the importance of the predictors.And feature extraction and model are performed based on this.This paper combines deep learning method with wireless propagation models to establish an intelligent wireless propagation model based on deep learning.Compared with traditional wireless propagation models,the establishment process is simpler and smarter.And there is no need for each scene to correspond to a propagation model.The historical data used in the model establishment saves additional labor and material costs.Compared with the multivariate linear regression model and the XGBoost model,the model in this paper has lower loss values on the train set and test set.And the prediction effect is better.In summary,the method of feature selection by the feature comprehensive index in this paper is effective.And this method can obtain useful features for model establishment.Moreover,the intelligent wireless propagation model based on deep learning proposed in this paper has certain feasibility.In practice,the prediction effect is better,which has great guiding significance and reference significance for the construction of 5G network base stations in China.