Online Tensor learning for Spatio-Temporal Financial Market Prediction

Financial market prediction is significantly important for investment decision making, but it is one of the most challenging tasks as the markets are affected by multiple factors. This paper aims to predict financial market data using spatio-temporal tensor model in online fashion with different spatial weight matrices. We introduce a new spatial weight matrix to represent the association of financial variables and compared ones based on location and correlation distance. Finally, a regression model is presented for multivariate financial data prediction. Models are tested and validated using data from Reuters and Factset database.