Research & Optimization Of Rough Set-Decision Tree Neural Network Forecast Model

Aiming at the problem of the low efficiency and accuracy of the neural network forecasting model,a rough decision tree algorithm and a particle swarm algorithm are proposed,which optimizes the neural network structure and parameters and improves the efficiency and accuracy of forecasting model.First,in order to solve the problem of large data and incremental data processing in static algorithm,a dynamic rule extraction algorithm based on rough decision tree is constructed.The rough set is combined with the decision tree by using the rough set to reduce the data dimension,and the sample is extracted by incremental method.After dynamic reduction,decision tree construction,rule extraction and selection,the matching cycle process realizes the data Dynamic Rule Extraction.Rough decision tree algorithm is used to determine the hidden layer of neural network,so that the hidden layer selection can be more scientific and rational,streamlining model structure.The algorithm can obtain more data implicit information and can be used in data mining.It provides a research method of online rule extraction.Second,in order to avoid the premature convergence of the network learning process to the local minimum,the particle swarm optimization algorithm is used to optimize the initial parameters of the BP network.Model of online optimization remains to be further studied.At last,The data of air quality and automobile evaluation in 2014 were collected.The rough decision tree algorithm and the particle swarm algorithm are combined to establish the neural network prediction model,and the training and simulation of the model was realized by MATLAB programming.Neural network with different hidden layer,and neural network prediction model which is not optimized,are analyzed and compared.The results show that the model can improve the accuracy and speed of prediction,which has theoretical value and popularization value.