Abstract:

In the light of the features of small sample size and some information unknown time series data, gray neural network was constructed by combining the gray theory with neural networks, which makes the full use of the advantage of the two kinds of approaches to realize the data mining of the small sample time series data effectively. Meanwhile, in order to improve the prediction accuracy of the model, the residuals were used for the model effective correction. The experiment results show that the proposed residual correction gray neural network has a high prediction accuracy, and is very suitable for the small sample time series data mining.

Key words: data mining, gray theory, neural networks, residual correction