Presented in this paper is a predicting model for time series forecasting of stock market index with the aid of bidirectional two-dimensional locality preserving projection and radial basis function neural network. First, 30 technical indicators in securities market are selected as the input features, and then adopts a sliding window to obtain the input data of the model. Next, by using two-dimensional locality preserving projection algorithm, reduction and feature extraction of the raw time series are performed in both horizontal and vertical direction. Finally, RBFNN is used to predict the closing price. Compared with the traditional dimension reduction methods, such as principal component analysis (PCA) and locality preserving projection (LPP), bidirectional two-dimensional locality preserving projection algorithm can be more effective to extract discriminating features. As verified on the Shanghai stock market index and the NASDAQ stock market index, the proposed model works well in ensuring the predicting accuracy which proves the effectiveness of this model in financial index time series prediction.