Coverage Estimation of Mobile Network Using Supervised Learning Model on Artificial Estimation Dataset

Minimization of driving test (MDT) is a key feature for network performance monitoring and optimization in mobile cellular networks. It has been widely used for network coverage estimation. However, the utility of MDT data envisages some challenges, especially positioning inaccuracy and data sparsity. This paper innovatively constructs a multi-source dataset, which is mainly based on MDT and combined with human empirical data. Then, we train a supervised learning model using Gradient Boosting Decision Tree (GBDT). Furthermore, the proposed model is used to estimate coverage for Scenario-based area in 31 cities, the result shows that the proposed model has practicability for telecom operator.