基于集成朴素贝叶斯模型的在线地图匹配方法

Map-matching based on ensemble learning of nave Bayesian models

为了实现移动对象数据库中准确的路网匹配,提出了基于历史数据训练多个朴素贝叶斯模型实现路网匹配的新方法。该方法针对不同天气、时间段和道路等级,从历史匹配轨迹中训练朴素贝叶斯模型。为了克服数据噪声和单个模型引起的匹配偏差,采用集成学习技术组合多个匹配器,通过投票实现高鲁棒性的地图匹配。在实际数据和模拟数据上的实验结果表明,该方法可以准确实现在线地图匹配。

To realize efficient map-matching in moving object database, train multiple naive Bayesian models is proposed to achieve this goal. For different weather, timestamps and road classes, the specific naive Bayesian models are trained. Further- more, to reduce the error caused by data noise and limitation of single hypothesis, ensemble learning is employed to train multiple matchers, thus the accuracy of map-matching is improved by robust majority-voting. Experiments on real and simulated data set show that the proposed approach can achieve map-matching with high precision.