基于自适应K值选择的K近邻算法研究

K Nearest Neighbor Algorithm Based on Adaptive K Value Selection

针对室内定位中基于位置指纹的K近邻法采用固态K值无法得到最优定位结果的问题,提出自适应K值选择的K近邻法。算法利用相邻定位点短时间间隔内空间位置变化引起的信号强度变化规律推测运动趋势,并与不同K值的定位结果构建的空间矢量进行匹配,从而自适应地从K近邻法的不同K值中选取最优的K值。同时依据室内AP的几何布局特征划分多个矢量域内,并对定位结果进行区域改正。试验结果表明,该算法能够很好地抑制较大误差的出现,提高定位的实时性、定位精度和稳定性。

The K nearest neighbor algorithm of location- based fingerprint with solid K value will not get the best result. Hence, a new K nearest neighbor algorithm through a daptive K value selection is proposed to resolve this issue. Algorithm uses the discipline that the change of signal strength during short intervals to speculate movement trend. And it matches with the space vector that constructed by the positioning results of different K. Thereby, the best K value can be adaptively chose from the different K values of K nea-rest neighbor algorithm. At the same time, based on the geometric layout features of indoor＇s AP , testing ground is divided into multiple vector filed, and correct positioning results by vector field. Experimental results show that the algorithm can suppress the occurrence of a large error, and iraprove real-time positioning, positioning accuracy and stability.