The Query Prediction And Optimization Of Graph Databases

In recent years,with the large increase in data-centric applications,the form of data has become increasingly complex and diverse.In addition to the traditional relational data model,the graph data model has gradually attracted attention.At the same time,the development of graph databases is also very rapid.For users,users are often more concerned about the efficiency of their use of the database.For this reason,database administrators need to optimize the index of the graph database to different degrees and to the users.Query optimization,in addition,you can also use the cache mechanism to optimize,the graph database often needs a higher time to read data from the disk,if you can predict the data that the user may query in advance and cache the data in memory,the system's response time to user queries can also be greatly improved.Existing query prediction algorithms and query optimization algorithms are lacking in cross-data portability and require high prices for queries and data.Therefore,this article studies SparQL's query prediction and query optimization.The main research contents are as follows:(1)For the query prediction of SparQL,this topic proposes a feature conversion method with reductive and more information,converts SparQL query into feature vector,and proposes a query prediction algorithm based on Seq2 Seq model,which The algorithm only uses the information of the query itself to make predictions,ensuring cross-data portability.The experimental results show the effectiveness of the algorithm;(2)For the query optimization of SparQL,in order to ensure cross-data portability and avoid statistical information errors,this topic regards the problem of triple connection order in query optimization as an enhanced learning task,and proposes a deep-based Q-Network's query prediction algorithm.Experiments show that the algorithm has better performance than Jena's query optimization.