High accuracy indoor positioning approach using kNN and LSTM algorithms

Abstract

In this paper, an effective approach to improve indoor positioning accuracy using machine learning is presented. The goal of the proposed solution is to reduce the distance estimation error by combining two algorithms k Nearest Neighbor (kNN) and Long Short-Term Memory (LSTM). Simulation results show that our solution achieves an accuracy of 40% when the required error is less than 1 meter, is higher than 26% and 14%, which respectively, of other studies using machine learning on the same data set and similar simulation scenarios.