In this study, an independent component analysis (ICA)-based signal reconstruction with neural network is proposed for financial time series forecasting. ICA is a novel statistical signal processing technique that was originally proposed to find the latent source signals from observed mixture signal without knowing any prior knowledge of the mixing mechanism. The proposed approach first uses ICA on the forecasting variables to generate the independent components (ICs). After identifying and removing the ICs containing the noise, the rest of the ICs are then used to reconstruct the forecasting variables. The reconstructed forecasting variables will contain less noise information and are served as the input variables of the back propagation neural network (BPN) to build the forecasting model. Experimental results on TAIEX (Taiwan Stock Exchange Capitalization Weighted Stock Index) closing cash index show that the proposed model outperforms the BPN model with non-filtered forecasting variables and random walk model.