Abstract
Machine learning is one of key-enablers in case of novel usage scenarios and adaptive behavior within next generation mobile networks. In this paper, it is examined how model-driven approach can be adopted to automatize machine learning tasks aiming mobile network data analysis. The framework is evaluated on classification task for purpose of base station anomaly detection relying on Neo4j graph database. According to the experiments performed on publicly available dataset, such approach shows promising results when it comes to both classification performance and reducing the time required for operations related to data import and model training.