室内定位特征选择算法研究-国内外oa导航.pdfVIP

软件 2015 年第 36 卷 第 1 期 SOFTWARE 国际IT 传媒品牌 室内定位特征选择算法研究 张乐玫 (北京邮电大学 信息与通信工程学院 北京 100876) : 如何能提供更高精度、更短预测时间、更少成本花费和更强实用性的应用是当下室内定位的研究热点。针对 提高定位精度，同时缩短定位预测时间的问题，本文将特征选择算法引入到传统的指纹定位法中，应用主成分分析法，将 终端收集的数据进行空间变换、特征压缩，从而去除数据噪声，在实测过程中达到更高定位精度，降低定位模型复杂度， 缩短预测时间的目的。同时，本文引入提升算法，通过结合并优化决策树算法，重点关注错分样本，达到优化模型训练过 程，增强模型实测泛化能力的目的。特别的，将提升算法与特征选择相结合，能够提供更精确、实用能力更强的定位方法。 : 模式识别与智能系统；提升算法；特征选择；室内定位 中图分类号: TP311 文献标识码: A DOI： 10.3969/j.issn.1003-6970.2015.01.008 本文著录格式：张乐玫. 室内定位特征选择算法研究[J]. 软件，2015，36(1)：3846 Study of feature selection algorithms in Indoor Positioning System ZHANG Le-mei (Beijing University of Posts and Telecommunications, Beijing 100876, China) 【Abstract 】: Nowadays, new needs are put forward for indoor positioning technology, such as higher location accuracy, less prediction time, smaller cost consumption and more practical applicability etc. In this paper, in order to improve indoor po- sitioning accuracy, real-time prediction ability, feature selection algorithm is introduced in traditional fingerprinting algo- rithms. PCA, Principle Component Analysis, is a feature selection algorithm by means of spatial transformation, characte- ristic compression of terminal collected dataset, in order to remove data noise, acquire high positioning accuracy reduce model complexity, shorten prediction time in practical positioning process. Meanwhile, this paper is introduced boosting algorithm into indoor positioning system to optimize model training process and enhance generalization ability within prac- tical prediction process by combining and optimizing decision tree algorithm, focusing on misclassification samples. Spe- cially, with the comb