Research On Data Graph Retrieval Algorithm Based On Bayesian Network

Data graph is a pattern which is constituted by a number of given points and lines connecting two points. With the points representing things and the lines representing corresponding relationship between two things, this pattern is usually used to describe a certain specific relationship among things. A number of query methods were proposed which were aimed at the deterministic data graph. However, there are still a lot of uncertainty information. To solve this problem, some scholars introduced probabilistic theory into the uncertain data graph, this provides a new method for querying and processing uncertain data graph, but the probability value is uncertain in some cases. For example, a disease induced by the probability of another disease, in the case of constant external conditions, this probability value is determined, but when the condition changes, this probability value can only be used as a reference.Bayesian network is a retrieval method works under the given model and rules, which has a unique advantage in the uncertainty reasoning. Due to the uncertainty of the data in disease areas, the traditional Bayesian network retrieval model can not meet the user’s search request Based on the study of traditional Bayesian network retrieval model, the Bayesian network data graph retrieval model is proposed which includes resource network and query network.The application of Bayesian network in uncertainty reasoning is introduced into data graph retrieval in order to solve the uncertainty problem encountered in the data graph retrieval, and an object graph method based on the conditional probability table is proposed.A global reasoning is needed when retrieved by Bayesian network, this practice requires a lot of time although it improves the precision rate. Since Bayesian network need much time, it has a disadvantage in the retrieval efficiency. Therefore, an effective subgraph retrieval algorithm is proposed in this thesis Bayesian method just acts as an aid in data graph retrieval. Reasoning will be done only when the indirect neighbor conditional probability is absent in this thesis, data graph structure information is still through efficient data retrieval algorithms to find out. On this basis, this thesis implements the data graph retrieval prototype system based on the Bayesian network, and experimental results showed that the effective subgraph retrieval algorithm has better precision.