Location recognition system using random forest

Abstract

There are many studies underway on indoor location recognition technologies, which use WiFi. BLE beacons, etc. This study proposes a location recognition system that works based on gathered WiFi data (BSSID, RSSI), which provides users with their location information. There are many multivariate analysis methods available, but random forest was used in this study, an ensemble learning method. The performances of differentiation were compared, applying random forest, an ensemble learning method of multivariate classification methods to the location information system. As for the experimental space, a limited interior space was divided in a grid shape to carry out an experiment. As a result of the experiment, the proposed algorithm had an accuracy, average 6.04% higher than the RSSI-based random forest location recognition, and the execution rate of the proposed algorithm, too, was consistent regardless of the number of the collected data, while that of the compared algorithm increased in proportion to the number of the collected data. The findings of our study are expected to be used in the areas of mobile advertisement, emergency rescue, and indoor travel guidance.