Forecasting Of Csi 300 Index Based On Pso-Lstm-Rt Composite Model

Abstract

This paper selects particle swarm optimization (PSO) to optimize the main parameters of LSTM and constructs PSO-LSTM model to forecast the CSI 300 index using quantitative factors. On this basis, considering the impact of textual factors such as the level of economic development, economic policy fluctuations and investor sentiment on stock index, a novel PSO-LSTM-RT composite forecasting model is proposed to analyze the common impact of quantitative and textual factors on stock index fluctuation. The empirical results showed that the forecast accuracy of PSO-LSTM is improved compared with BPNN and LSTM, while the PSO-LSTM-RT further improved the accuracy by 0.93%. This paper concludes that the PSO-LSTM-RT model which takes into account the quantitative and textual factors has better forecasting performance.