Indoor Localization using Machine Learning and Beacons

Abstract

In this paper, we use beacons and machine learning to localize indoor positions. The data used for machine learning consists of the RSSI value received by smartphones with eight beacons and the numerical code value, which means 13 indoor zones. K-Nearest Neighbors algorithm is used for model training. The original data is refined into two data that have a label as detailed space and approximate space, and the models train for two data. Training results show that the models achieve high accuracy for both datasets. As a general idea, Models are more accurate when training with data whose labels are approximate spaces.