Research On Prediction Of Trajectories Of Moving Objects Based On Historical Information

With the rapid development of information technology, the explosive growth of data and information has become the common challenges and opportunities for all walks of life. Thus, the information society has entered the era of big data. The Big data not only changed people’s production and life-style, but also fundamentally changed the ways of scientific research. The trajectorie of moving object contains a lot of information,such as location and behaviorwhich has a wealth of features of motion about different moving objects.By excavating and utilizing trajectory data, we can not only understand the commonalities of generic behavior of different moving objects, but also grasp the characteristics. Therefore, moving trajectory analysis has been widely concerned by scholars at home and abroad. As an important branch of the moving trajectory analysis, Moving objects trajectory prediction has significant application value in various fields andbroad application prospect in the actual life. In this paper, related algorithm and its application has been studied for the trajectory prediction of different moving objects. The main work is as follows.1. Studying on the principle and applications of trajectory prediction, such as BP (Back Propagation) neural network, SVM (Support Vector Machine) classification and Markov chain model.2. Appling trajectory prediction technology to the underground safety protection system. Due to the complicated environment under the shaftthat, sometimes, the anchor node invalidswill result in some data anomalies or missing, so that data transmission ability of moving objects underground is cut down. Therefore, we propose a prediction model using k-nearest neighbor and BP neural network. The k-nearest neighbor algorithm is used to preprocess the abnormal or missing data. Then, the preprocessed data is used for the final prediction. The experimental results show that the prediction accuracy is improved.3. Proposing an improved Subsyn algorithm for taxi destination prediction. In order to improve the quality of LBS (Location-based Service), the prediction model needs to predict the destination of trajectory of moving objects as accurately as possible. Mostly existing methods about destination prediction of trajectory obtains the prediction result by entering the historical data to the prediction model. This method deals with timestamp as a subsidiary dimension, which can not fully reflect the characteristic value of the forecast target. In this paper, we add track ID, taxi ID, taxis original Location ID, timestamp and holiday details as data characteristic values to the process of training SVM. However, these data characteristic values including in the traditional first-rank Markov chain model increase the time and space complexity. Therefore, we propose an approachfof destination prediction by improving Subsyn algorithm, the algorithm obtain the posterior probability by training SVM model on trajectory data. Then, combining the posterior probability with Bayesian probability, the final predicted value of maximum probability is given by weighting. Lastly, the algorithm completed the destination prediction of moving objects.In summary, the innovations are as follows.1. Using the improved BP neural network algorithm in underground target trajectory prediction, the k-nearest neighbor algorithm is used in the preprocess about missing data. And then, a new data set is constructed by the pretreatment data. Finally, the new dataset is used for training BP neural network.2. An improved Subsyn algorithm was proposed to predict the destination of the taxi track. Basing on Subsyn algorithm, the proposed algorithm calculates the posterior probability of SVM classification. Finally, the maximum predicted value is calculated by weighting.