A WEIGHTED NON-LINEAR GREY BERNOULLI MODEL FOR FORECASTINGNON-LINEAR ECONOMICTIME SERIES WITH SMALL DATA SETS

作者：Wang, Zheng-Xin^*

来源：Economic Computation and Economic Cybernetics Studies and Research, 2017, 51(1): 169-186.

• grey systems theory
• time series prediction
• non-linear grey Bernoulli model
• small data sets
• stock exchange composite index

To accurately forecast non-linear economic time series with small data sets, the weighted non-linear grey Bernoulli model (WNGBM) is built in this paper. Through the optimization of the power index and weights for accumulative generation, WNGBM can more actively adapt to non-linear fluctuations in the raw data than NGBM. A typical case of topological rolling prediction verifies that the WNGBM exhibits better non-linear prediction capabilities than other grey models. Furthermore, the forecasting performance of WNGBM is compared with that of Holt-Winters, Support Vector Regression (SVR), and BP Neural Network (BPNN) based on the Shanghai Stock Exchange(SSE) Composite Indices. Results indicate that WNGBM shows the best ability to fit non-linear data from small sample sizes, while it has a slightly higher error in the prediction of out-of-sample data for the SSE Composite Index than that of BPNN. The extreme values mean that the prediction curve of the Holt-Winters method generally deviates from the actual data, which leads to a greater prediction error.

• 出版日期2017
• 单位浙江财经大学