A Comparative Analysis of Regression Based Indoor Localization Techniques

Indoor localization is essential for numerous applications where position of a sensor node is a prerequisite. In past few years, several techniques have been proposed for the estimation of location of a sensor node. Currently focus of researchers has been shifted towards machine learning in order to find solutions for problem of indoor localization. In this paper, few machine learning based regression algorithms have been investigated for their suitability to localization problem. K-Nearest Neighbor (KNN), Decision Tree (DT) and Forest Tree (FT) regression algorithms were explored and utilized to generate models for prediction of location of a sensor node. An empirical data was used for the experiments. The experimental results showed that KNN regression model compromised with localization accuracy but DT and RF depict accurate localization outcomes. Statistical analysis proved that DT and RF performed better than KNN regression model.