Abstract: In order to solve the problem that the traditional clustering algorithm is easy to fall into the local optimum and affect the positioning accuracy, an improved iterative self-organizing data analysis clustering algorithm for indoor positioning of Wi-Fi is proposed. In the offline stage, the Euclidean distance standard deviation of each point in the fingerprint database was calculated, the initial parameter threshold was optimized, and the clustering center was dynamically selected to reduce the location error. In the online stage, the adaptive weighted K-nearest neighbor algorithm was combined with the clustering algorithm to avoid the influence of fixed K value on the positioning results and effectively improve the positioning accuracy. The improved algorithm was applied to an engineering example for verification. The results show that the proposed algorithm has a probability of 63.33% when the positioning accuracy is within 1 m and a probability of 90.00% when the positioning accuracy is within 2 m, which verifies the effectiveness of the proposed algorithm.