Using Semi-supervised Machine Learning with partially labeled Data for Indoor Outdoor Detection
11 January 2019
This paper demonstrates the feasibility of an hybrid/semi-supervised classification method for detecting the environment of an active mobile phone based on cellular radio labelled and unlabelled data. Precisely, we provide answers, with minimal intervention of mobile users, to the following questions: what is the environment of the mobile user when it is/was experiencing a mobile service/application: is it indoor or outdoor? The method implementing at the mobile network side is interesting for mobile operators since it is low complexity, less human intrusive and accurate. The semi-supervised classification algorithm identifies the environment using large real 3GPP signals measurements collected within network. We empirically validate the innovative algorithm using real-time radio measurements, with partial ground truth information, gathered daily, weekly, monthly, for indoor and outdoor locations and in multiple typical and diversified environment crossed by mobile users. The study confirms the effectiveness of the proposed scheme compared to the existing classification methods.
