Topology Regressive Distributed Model for Financial Time Series Prediction

Financial time series prediction has long been one of popular research areas for monetary policy makers, market participants and researchers. Inspired from the principle of self-organizing map, the author propose a topological regressive distributed model, which inherits the topology reserving advantage of self-organizing map and the simplicity of the normal autoregressive model. Comparison studies show the proposed method outperforms neural network based local model approaches and traditional autoregressive model on non-stationary financial time series (e.g. exchange rate, stock price).