Two-layer architecture of telco churn auto-ML

Abstract

In recent years application of machine learning has been used in many businesses and many different use cases. This led to increased need for machine learning applications. Creating machine learning applications is time consuming and requires expert knowledge of machine learning and domain expertise. This caused a development of the new research topic of Automated Machine Learning (AutoML). This research uses real telecom data to check whether the AutoML can successfully predict customer churn. In addition this research proposes the two-layer architecture of AutoML telco systems for self repairing/upgrading the model. In this work we propose the excluding the one period of implementation window, which is enabled by the use of AutoML.