Application of singular spectrum analysis and kernel-based extreme learning machine for stock price prediction

Stock prediction is known as one of the most challenging task in the field of time series analysis. Many existing studies emphasize only in improving accuracy, but not many are concerned about improving speed of the prediction. This paper proposes a stock prediction model using kernel-based extreme learning machine with singular spectrum analysis (SSA-KELM) as a preprocessing tool. The prediction performance of SSA-KELM is compared to five other models. Three stock price data are used to evaluate the performance. The experimental results show that all SSA-based models can outperform non-SSA models. The results are also shown that SSA-KELM can achieve the highest accuracy and the lowest training time among other SSA-based models. The proposed model can therefore be considered as an efficient model for stock price prediction.