A Big Data-Driven Hybrid Solution to the Indoor-Outdoor Detection Problem

Abstract

Context-aware mobile applications are emerging as a relevant technology to improve the users' satisfaction. By being aware of the current situation of a user, a context-aware application can adapt itself to provide the most suitable services at that moment. Companies from different areas are investing in turning their applications aware of the users' context with the objective of increasing their quality-of-experience and reducing churn. Among many possible contexts, the situation of a mobile user in terms of the type of environment (i.e., indoor or outdoor) is a relevant yet difficult to obtain information. In this work, we propose the HybridIO , a solution to the Indoor-Outdoor Detection Problem that advances the state-of-the-art, since it requires less sensor data to operate and generalizes well in other domains. We validate our solution considering a large amount of real data, and the results reveal an improvement of up to 14% in precision when compared to other baselines. We also deploy the proposed solution into a real-time big data architecture that is able to enrich up to 400 records/second with the indoor-outdoor information.