In this paper, we present a new method for improving the eciency of indoor localization algorithms, in terms of running time and error rate, using the KD-tree data structure. One of the main challenges of indoor localization algorithms in large environments is the high processing overhead of these algorithms due to the high volume of input data and lack of processing resources in users' mobile devices. In the proposed method in this paper, we rst cluster the ngerprint database. Then, with the help of a newly proposed method, a modi ed KD-tree is implemented according to the conditions of the clusters. This tree is a decision-making structure to select one speci c cluster where the user stands there. Finally, when a user entered, using a few simple comparisons in the KD-tree, the desired cluster is found and only information about that cluster is passed to the localization algorithm, to compare and predict the user's location. The results of the implementation of this method on the ngerprint data set of the Faculty of Engineering at Arak University show that the proposed method reduces the running times and errors to less than half the values, compared to the time of not using the proposed method.
