改进K-means聚类的自适应加权K近邻指纹定位算法

Adaptive weighted K-nearest neighbor fingerprint location algorithm based on improved K-means clustering

针对指纹定位精度易受指纹数据K-means聚类预处理效果不佳、加权K近邻算法采用固定K值进行匹配定位精度差等问题,提出一种基于改进K-means聚类的自适应加权K近邻算法。算法在对指纹数据进行聚类计算过程中充分考虑参考点间接收信号强度值与实际物理坐标的双重影响,以避免参考点分类不明确;根据每个测试点的匹配参考点之间实际距离的均值和标准差设置阈值,动态选择K值。实验结果证明,改进K-means聚类的自适应加权K近邻算法相较于传统室内定位算法定位精度提高了44%,可为相关应用提供更精确的定位服务。

An adaptive weighted K-nearest neighbor algorithm based on improved K-means clustering is proposed to solve the problem that the fingerprint positioning accuracy is affected by the poor K-means clustering preprocessing effect of fingerprint data and the fixed K value used in the weighted K-nearest neighbor algorithm for matching positioning.In order to avoid the occurrence of ambiguous classification of reference points,in the process of clustering fingerprint data,the algorithm fully considers the double effects of the received signal strength between reference points and their actual physical coordinates.K values is dynamically selected according to the mean value and standard deviation of the actual distance between the matching reference points of each test point.The experimental results prove that the improved adaptive weighted K-nearest neighbor algorithm for K-means clustering can provide more accurate positioning services for related applications.The positioning accuracy is improved by 44%compared with the traditional indoor positioning algorithm.