The Arc Learning Algorithm Based on Extended Functional Dependency

Taking the consideration about that the existing structure learning methods for Credal Network Structure separate the relation between arc’s existence and direction and require sample data strictly, the concept of extended functional dependency is put forward to solve the above problem; Make full use of extended functional dependency to express the existence and direction of the arc, so it makes the two processes united. Furthermore, the concept of dependency degree is brought forward to make the requirement for data set not so strictly; thus, the available of the network structure in the context of incomplete data set is ensured. Transfer the question of structure learning into the question of extended functional dependency discovering. The algorithm for discovering extended functional dependency is designed, and the accuracy of the method is proved with examples.