WiFi-Based Indoor Positioning by Random Forest and Adjusted Cosine Similarity

WiFi based indoor positioning technology is very popular due to no additional hardware. WiFi location fingerprint is widely used for WiFi-based indoor positioning system. They are collected into fingerprint database to be used for estimating the location. In order to accurately estimate the indoor location, a kind of improved random forest algorithm is proposed in this paper. The proposed algorithm utilizes regional grid division to reduce the maximum error, and adopts adjusted cosine similarity to match the proper grid and fingerprint. Compared to the original random forest, the proposed algorithm reduces 1.15m in terms of maximum error.