The positioning system based on WIFI location fingerprint can achieve high precision indoor location. The neighbor selection algorithm based on Received Signal Strength Indicator (RSSI) is easy to introduce singular points when locating indoors, which leads to the decrease of positioning accuracy. To solve this problem, this paper proposes a Similarity-based K-Nearest Neighborhood Location Algorithm (SKNN). Referring to the idea used to solve the problem of similarity of nodes in bipartite networks, this algorithm builds a bipartite network between the location fingerprint and the AP. It proposes a similarity parameter which can be used to modify the K-Nearest Neighborhood localization algorithm. The experimental results show that the SKNN algorithm proposed in this paper can effectively reduce the influence of singular points on the positioning results and improve the positioning accuracy, with 80% of the positioning errors within 2m, and the effect is obvious in the large scene.