Ship Collision Risk Accident Analysis And Anti-collision Decision Making Method Using Text Mining

With the proposed national strategies of “the Belt and Road” and “the Yangtze River Economic Zone”,the development of water transportation on the main line of Yangtze river,which crossing the east,middle and west of China,is facing new opportunities and challenges.At present,the maritime traffic safety situation is still grim,and water traffic accidents happen frequently.Among them,ship collision accidents account for the largest proportion in all water traffic accidents.In order to improve the safety of traffic on the main line of Yangtze River,this thesis carries out the research on analysis and control measures of ship collision accidents based on text mining.Firstly,the urlopen function of R language is used as a means of text mining corpus acquisition to achieve automatic crawling of accident or distress reports on the official website of maritime authority.The human factors,ship factors,environment factors and management factors in the corpus will be used as the target data.Invoking the two language feature selection packages—“Bruta” and “Caret” in R language,a high dimensional sparse set of original eigenvector space is obtained,which contains the feature items and their weights.Using the chi-squared statistic to reduce the dimension,33 text feature items that represent the ship collision risk factors are acquired.And the wordcloud package in R language is used for date visualization.Secondly,based on the ship collision risk factors by text mining,combined with previous studies,a prior distribution is introduced.Taking voyage preparation,navigational vigilance,collision avoidance decision and manipulation four steps of the collision process as the main line,a ship collision risk Bayesian network structure based on “human-ship-environment-management” system is conducted.Combined with the ship collision accident or distress reports,the conditional probability table of each node in the Bayesian network structure is calculated.So the ship collision risk model can be established in the Hugin Lite8.0 software.Through a case study,it is verified that the probability of collision is more than 90% under the circumstance of relevant conditions.And the primary cause of ship collision accidents is released by this model.Finally,aiming at the ship collision factors which obtained from the text mining,combined with the traditional rule of ship collision avoidance,prevention and control countermeasures of ship collision risk are proposed with the use of simulation from two angles of time and space,which conform to the collision avoidance rules of early,large,wide,and clear.A direct collision risk prevention and control strategy is provided for inland crews.Based on collision avoidance rules,prevention and control countermeasures of single ship's and between two ships' collision risk are studied by the method of ship motion simulation.A case is used to verify the countermeasures for collision risk prevention and control of inland ships.