基于网格概率的离群点检测算法

随着移动网络、智能终端的迅猛发展,基于位置的服务LBS(Location-based Service)越来越热门,因此基站位置信息的正确与否成为关注的重点.针对基站地理位置存在部分错误这一现象,提出了基于网格概率的离群点检测算法来核查错误的基站.首先,根据基站分布的规则将数据空间分成若干网格单元;其次,根据用户轨迹签到信息关联出其在动态时间范围内经过的基站序列,将基站序列映射到网格中,计算出临近网格单元集合;最后,根据基站分布特点对网格单元内目标基站的临近基站求隶属概率,筛选出离群点,即错误的基站.实验表明,该算法的时间复杂度低且核实准确率较高.

Outlier Detection Algorithm Based on Grid Probability

With the rapid development of the mobile networks and intelligent terminals, location-based service has become more and more hotter on the internet, therefore the correction of the base stations'' position becomes a critical factor. For the wrong base stations are uncertain, it proposes a new detecting algorithm based on the probability of the near grids, which is used to verify the wrong base stations. Firstly, it divides the data space into some grids. Secondly, combining with the users'' attendance location information, it gets the track of the base stations in a short dynamic time and maps them to the corresponding grids. Finally, referring to the position characteristics of the base stations, it could give the membership probabilistic and filter the outliers, that are the wrong base stations. The results show that the algorithm has low complexity and high accuracy of detecting the wrong ones.