摘要
针对传统轨迹预测方法在历史轨迹数目有限时,预测准确度较低的问题,提出一种改进的贝叶斯推理(MBI)方法,MBI构建了马尔可夫模型来量化相邻位置的相关性,并通过对历史轨迹进行分解来获得更准确的马尔可夫模型,最后得到改进的贝叶斯推理公式。实验结果表明,MBI方法比现有方法的预测速度快2到3倍,并且有较高的准确度和稳定性。MBI方法充分利用现有轨迹信息,不仅提高了查询效率,还保证了较高的预测精度。
The existing algorithms for trajectory prediction have very low prediction accuracy when there are a limited number of available trajectories. To address this problem, the Modified Bayesian Inference （MBI） approach was proposed, which constructed the Markov model to quantify the correlation between adjacent locations. MBI decomposed historical trajectories into sub-trajectories to get more precise Markov model and the probability formula of Bayesian inference was obtained. The experimental results based on real datasets show that MBI approach is two to three times faster than the existing algorithm, and it has higher prediction accuracy and stability. MBI makes full use of the available trajectories and improves the efficiency and accuracy for the prediction of trajectory.
出处
《计算机应用》
CSCD
北大核心
2013年第7期1960-1963,共4页
journal of Computer Applications
基金
国家重大科技专项(2011ZX02507-006)
关键词
轨迹预测
马尔可夫模型
贝叶斯推理
trajectory prediction Markov model Bayesian inference
