Data-Driven Mobile Network User Satisfaction Analysis: An Optimization Method Based on Multi-Model Fusion

Abstract

With the rapid development of mobile communication technology, improving user satisfaction has become a core focus for mobile network operators. In an era of ubiquitous connectivity and homogeneous services, understanding and enhancing user experience is crucial for maintaining competitive advantages. Traditional methods to improve user satisfaction, such as responding to complaints and resolving specific issues, have become insufficient with the expansion of user bases and the diversification of mobile services. This study proposes a data-driven approach to analyze the factors influencing mobile user satisfaction in Beijing, focusing on voice and internet services. We combine feature engineering, decision tree models, and ensemble learning techniques to predict and quantify user satisfaction. Specifically, we use the CART (Classification and Regression Tree) model to extract feature importance, and integrate Random Forest and XGBoost models to further improve prediction accuracy through hyperparameter optimization and model fusion. This method combines structured data with unstructured text data, including user descriptions and service notes, to deeply explore the core factors of user experience. Experimental results show that multi-model fusion significantly improves prediction accuracy, with factors such as GPRS traffic, monthly usage, and network issues identified as the main drivers affecting user satisfaction. This study provides valuable insights for mobile network operators to optimize services and enhance customer experience. Additionally, the proposed method can be extended to other regions and applied to multiple industries where customer satisfaction is critical.