Modified Bayesian Inference for Trajectory Prediction of Network Virtual Identity Based on Spark
Yuhang Song, Gang He, Dechen Yu · 2016
Traditional trajectory prediction method shows low prediction accuracy when dealing with a limited number of historical trajectory. Therefore an improved Bayesian inference (MBI) method is proposed, which builds a Markov model to quantify the correlation between adjacent locations, and provides a more accurate Markov model by decomposing the historical trajectory. Finally we improve Bayesian inference formula. The experiment uses Spark platform processing original data and intermediating results of calculations as the large amount of data to be calculated. The results show that, MBI method is 2-3 times faster than the speed of existing prediction methods, and has higher accuracy and stability. MBI method makes full use of existing track information, not only to improve the query efficiency, but also to ensure a higher prediction accuracy.