Research and Improvement of Splitting Rule Extraction Data Mining Algorithm Based on Neural Networks

Neural networks can make higher classification accuracy in the problem of rule extraction, for it has the capability of non-linear mapping and exact approaching mechanism. This paper proposes the procedures of a kind of splitting rule extraction algorithm based on feed forward network. After advancing the input nodes within this complex architecture, build up a RBF (Radial Basis Function) neural network classifier using JOONE (Java Object Oriented Neural Network) to optimize the inputs. Then the complexity of rule extracting procedures will be reduced and simplified. And the experiments result shows the validation and effectiveness of this improving method.