Sparse Representation Based Approach to Prediction for Economic Time Series

This paper addresses the problem of economic time series forecasting, and a new prediction method is proposed. The method fully capitalizes on the two key technologies, sparse representation, and fuzzy set theory, to handle the stock time series forecasting problems. First, sparse representation is applied to smooth the time series. Then, the fuzzy technology is used to convert time series into fuzzy series. Based on the number of the occurrence of fuzzy sets, the weights are calculated. Finally, the future value of the time series can be forecasted using the weights and inverse transformation of sparse representation. The experimental results show that the proposed method produces more accurate forecasted results.