Research On Trajectory Prediction Method Based On Deep Learning

Today's world is gradually moving into the field of big data.In daily life,people have more ways to get data,which is more convenient than before.These data also provide solutions to some important questions.A typical application scenario is to avoid collisions in maritime traffic.In the past period of time,in response to this problem,people have proposed many methods of processing and mining trajectory data,which has promoted the rapid development of the field of trajectory processing.The rise of artificial intelligence has further improved the safety and efficiency of data processing.After summarizing the research status of trajectory data processing,this paper proposes a feasible method to predict maritime traffic trajectory by processing AIS data based on neural network,which can effectively increase the safety of marine operations and prevent the hidden dangers of close encounters.The purpose of this method is to accurately predict the future trajectory of the selected ship and provide an estimate of uncertainty relative to the predicted position.Through data preprocessing,data completion based on trajectory clustering,and trajectory prediction based on LSTM neural network,this paper improves the quality of the ship's AIS trajectory data and analyzes it.The key job of this article includes the following three parts:First,collect and sort the original AIS data set,and then perform preprocessing,mainly performing operations such as abnormal point processing,noise filtering,and normalization to improve data quality;In order to ensure a better prediction and filtering effect,take the actual ship's AIS trajectory data as an example,use various algorithms to cluster,and then determine the sparseness of the trajectory,and perform data on the required trajectory according to the center trajectory of the same category.Completion.This paper studies and analyzes several commonly used clustering algorithms,such as K-means algorithm and DBSCAN algorithm,and compares the pros and cons of each algorithm and applicable scenarios;Finally,the LSTM neural network model is established,and the AIS trajectory data is used to continue the research.The trajectory after the completion is modeled using prediction algorithms such as LSTM recurrent neural network,and the trajectory prediction is performed after selecting the parameters.The final consequent proves that the prediction accuracy based on the supplemented data is significantly improved.