Aiming at characteristics of received signal strength（ RSS）

such as volatility

susceptibility to interference and problem that it is easy to be disturbed by the remote reference point（ RP） in the algorithm matching process

and resulting in a large deviation of the finalized positioning results

a fingerprint localization algorithm based on clustering results is proposed. Different from traditional clustering by reference point signal strength

K-means

++

algorithm by reference points location clustering is used. Select M APs（ max mean

MM）with the largest RSS mean

and using the supervised learning algorithm

KNN

to classify the test points

location coordinates can be calculated by using the deterministic matching algorithm weighted K-nearest nei

ghbor（ WKNN）. The experimental results show that the proposed method greatly improve the localization characteristics than the method of clustering by signal strength.