Ontology Semantic Based Modeling,Query And Reasoning For Mobile Trajectory

With the rapid development of Internet technology and significant improvement of data storage capacity,more and more datasets of vehicle mobile trajectory are collected.How to analyze the rule of vehicle mobile trajectory in dataset is a difficulty.There are some methods to analyze the rule of mobile trajectory for trajectory dataset.At present,the machine learning method is widely used to predict mobile position of the vehicle,but this kind of method has some shortages.For example,the machine learning method is used to predict the vehicle trajectory,which is easily affected by training set.If missing or inaccurate data is existed in dataset,this circumstance may lead to deviation of the prediction results,which couldn’t reflect the rule of vehicle mobile trajectory well.How to reduce the impact of data on prediction results is a problem to be solved in the field of mobile computing.In order to solve the problem of low prediction accuracy caused by incomplete data and other factors,ontology knowledge base is introduced into the vehicle trajectory for mobile prediction.This dissertation proposes an ontology-based model,which combined with Markov logic network(MLN)to predict the mobile position of vehicle so that prediction performance can be improved.The specific contents of this dissertation are as follows:(1)For spatio-temporal data,the longitude and latitude coordinates are transformed into semantic locations to improve the interpretability of the trajectory data.Considering the location relationship between roads and the location relationship of vehicle movement,the corresponding semantic rules are defined,and an ontology model of vehicle movement scene with constraints is built.(2)For incomplete trajectory data,the problem of incomplete information reasoning is studied.Taking the ontology model of vehicle movement scene as knowledge base,semantic query method is used to construct Markov logic network,and semantic query results are taken as training dataset to avoid the long reasoning time due to the large training dataset of MLN model.(3)The precision of second-order Markov chain,DS evidence theory and ontology-based Markov logic network reasoning methods are compared respectively.In addition,secondorder Markov chain,DS evidence theory and MLN are combined to verify whether the weight parameters affect the reasoning performance of Markov logic network.Experiment results show that Markov logic network based on ontology has better prediction results on two test datasets in different periods.At the same time,it is also verified that the reasoning precision of Markov logic network is related to the initial weight value of training dataset.And Markov logic network whose weight value is evidence theory has better prediction results.