基于人工神经网络的VoLTE MOS预测

VoLTE MOS Prediction Based on Artificial Neural Network

在现阶段的网优工作中,难以得到全网范围的VoLTE MOS数据,提出的基于人工神经网络的VoLTE MOS预测,采用MCMC方法对路测未能获取的缺失数据进行数据模拟以增强样本数据,对增强后的数据采用人工神经网络模型进行拟合并建立预测模型,进而得出VoLTE MOS的预测值,经理论验证以及与仪表实测结果进行对比,证实该方法有效可行。

In the network optimization at this stage, it is difficult to have the VoLTE MOS data in the entire network. A prediction about VoLTE MOS based on artificial neural network （ANN） is proposed. Specifically, the MCMC method is used to simulate the lost data on the drive test to enhance the sample data. Then, the ANN model is adopted to fit the enhanced data, build the prediction model and derive the predicted value of VoLTE MOS. By comparing the theoretical derivation and measured results, the proposed method is effective and feasible.