Adaptive K-nearest neighbour algorithm for WiFi fingerprint positioning

K-nearest neighbour is one of the most widely used algorithms for indoor positioning systems. However, the error for each estimated position notably varies depending on the K value used for the algorithm. Therefore, if K is a fixed value, the estimation error for the positions cannot be further reduced. In this Letter, I propose an algorithm that adapts the K value for each position by analysing the correlation between the K value and the received WiFi signal strength. The proposed algorithm provides an improvement above 30% on the positioning accuracy compared to the algorithm with fixed K value.