一种基于CFSFDP改进算法的重要地点识别方法研究
发布时间：2018-12-09 19:01
【摘要】：为解决CFSFDP聚类算法由于无法自动选择簇中心点而难以应用于重要地点识别的问题,引入一种簇中心点自动选择策略对算法进行改进。该策略将簇中心点权值的变化趋势作为自动划分簇中心的依据,有效避免了通过决策图判决簇中心点的方法所带来的误差。将CFSFDP改进算法与数据预处理及逆向地理编码等技术结合起来,能够以较高的精度实现重要地点识别。实验以Foursquare数据为例,结果表明CFSFDP改进算法比DBSCAN具有更高的准确率和较低的计算量,进一步证明了该方法在处理稀疏位置数据的重要地点识别问题上具有一定优越性。
[Abstract]:In order to solve the problem that CFSFDP clustering algorithm is difficult to be applied to the identification of important points because of its inability to automatically select cluster center points, an automatic selection strategy for cluster center points is introduced to improve the algorithm. The strategy takes the change trend of weight value of cluster center as the basis of automatic classification of cluster center, and effectively avoids the error caused by the method of deciding cluster center point by decision graph. By combining the improved CFSFDP algorithm with data preprocessing and reverse geographic coding techniques, important location recognition can be realized with high accuracy. Taking Foursquare data as an example, the results show that the improved CFSFDP algorithm has higher accuracy and lower computational complexity than DBSCAN, and further proves that this method has some advantages in dealing with the important location identification problem of sparse position data.
【作者单位】： 电子工程学院;
【基金】：国防重点实验室基金资助项目(9140C130104)
【分类号】：TP311.13
，
本文编号：2369879
[Abstract]:In order to solve the problem that CFSFDP clustering algorithm is difficult to be applied to the identification of important points because of its inability to automatically select cluster center points, an automatic selection strategy for cluster center points is introduced to improve the algorithm. The strategy takes the change trend of weight value of cluster center as the basis of automatic classification of cluster center, and effectively avoids the error caused by the method of deciding cluster center point by decision graph. By combining the improved CFSFDP algorithm with data preprocessing and reverse geographic coding techniques, important location recognition can be realized with high accuracy. Taking Foursquare data as an example, the results show that the improved CFSFDP algorithm has higher accuracy and lower computational complexity than DBSCAN, and further proves that this method has some advantages in dealing with the important location identification problem of sparse position data.
【作者单位】： 电子工程学院;
【基金】：国防重点实验室基金资助项目(9140C130104)
【分类号】：TP311.13
，
本文编号：2369879
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2369879.html
