Research Of Weapon Recommendation Method Based On Knowledge Graph

With the development of modern warfare and the improvement of tactics,combat missions are gradually refined,and the type of weapons is increasing.It’s a great challenge for commanders to find the potential weapons which are suitable for the combat missions based on excessive information.In order to help commanders improve finding efficiency,the weapon recommendation system develops rapidly.However,the sparsity and timeliness of traditional collaborative filtering algorithms limit the performance of the recommendation system.This thesis introduces the representation learning of knowledge graph and the time decay function into the collaborative filtering algorithm to improve the performance of the recommendation system.This thesis takes the weapon recommendation method based on knowledge graph as the research topic,focuses on the research of the knowledge graph representation learning,the weapon recommendation technology based on knowledge graph,and the implementation of weapon recommendation system.The main research work of this thesis is as follows:Firstly,this thesis presents the Trans R model with entity information which is a new representation learning method of knowledge graph used in weapon field.This method,which draws on the idea of Laplacian Eigenmap,integrates entity attribute information and entity structure information into the Trans R model in the form of similarity to improve the problem caused by the training mode which is based on triples.Through the experiments of link prediction and triple classification,it is proved that this method can effectively improve the performance of the Trans R model when it is used in the weapon knowledge graph.Secondly,this thesis presents a collaborative filtering algorithm which is based on the knowledge graph and time weight in order to solve the sparsity and timeliness of traditional collaborative filtering algorithms.Combining the knowledge graph representation learning,time decay function and collaborative filtering algorithm,using the usage records of weapons and the semantic information of weapons extracted from the knowledge graph,it can multiple-angle describe the weapon similarity,highlight the importance of recent usage records and make up for the defects that the item-based collaborative filtering algorithm only consider the usage information of weapons.The experiment proves that the method used in this thesis is effective.Finally,this thesis implements a weapon recommendation system which uses the collaborative filtering algorithm based on knowledge graph and the time weight.This system takes the combat mission selected by the user as input,and then predicts suitable weapons according to the recommendation algorithm and the usage records of weapons.The potential weapons which are suitable for the selected combat mission will be shown to the commanders for reference at last.