An adaptive selection decomposition hybrid model for stock time series forecasting

Stock price prediction is a hot topic that has drawn tremendous interest from scholars around the world. This paper aims to improve the prediction accuracy of stock time series. To this end, we design a novel Adaptive Selection Decomposition Hybrid (ASDH) model which can automatically search for the decomposition mode and the combination configuration based on BP Neural Network Optimized by Genetic Algorithm (GABP), K-nearest neighbor (KNN) and Autoregressive moving average model (ARMA). In addition, Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and Ensemble Empirical Mode Decomposition (EEMD) are adopted to decompose time series, and we combine Permutation Entropy (PE) and Hierarchical Agglomerative Clustering (HAC) to reconstruct time series. To prove the validity of the proposed model, we conduct a series of experiments on the closing price of four financial data, and compare the proposed model with other benchmark models. Finally, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percent Error (MAPE), Pearson correlation coefficient and the Model Confidence Set (MCS) test, serving as performance evaluation means, intuitively and effectively reflect the superiority of the proposed model in this paper.