Predicting stock trend using multi-objective diversified Echo State Network

The prediction of stock price movements is considered as a challenging task for financial time series analysis. The difficulty of predicting the trends lies in the dynamic temporality and noise in the stock data. The Echo State Network (ESN) is a popular time series prediction model that considers the temporality of the stock time series, but ESN often falls into the dilemma of over-fitting due to the existence of many unnecessary neurons in the hidden layer. To predict the stock price movements, in this paper, we propose Multi-objective Diversified Echo State Network (MODESN). MODESN assigns a diversity requirement to the ESN such that the network is able to maintain better generalization ability. Experiments are performed with Shanghai Composite index to evaluate the effectiveness of the proposed approach. Compared with classical Echo State Network, the proposed MODESN usually achieves higher accuracy and avoids over-fitting. The experimental results indicate that MODESN performs favorably compared with alternative methods.