Year: 2024
Type: conference-paper
Abstract: In recent years, financial markets have attracted the attention of many investors due to their great volatility and their considerable return on investment. Recently, it has been hoped to build some algorithmic models to make possible predictions about the financial markets. CSI300 is composed of the 300 most representative securities in the two largest exchanges in mainland China, Shanghai and Shenzhen exchanges. BTC is a decentralised, anonymously traded, non-homogeneous token with a total volume of 21 million coins. This paper predicts six metrics, IC, ICIR, RankIC, RankICIR, AR and IR for CSI300 and BTC based on the Informer model. In this paper, the Encoder layer of the Informer model is changed to a three-module stacking process. The prediction results of the final Informer model for CSI300 are 0.051 ± 0.001 for IC, 0.39 ± 0.04 for ICIR, 0.051 ± 0.002 for RankIC, 0.43 ± 0.04 for RankICIR, 0.22 ± 0.06 for AR, and 2.0 ± 0.4 for IR. The prediction results of the final Informer model predicts BTC with IC of 0.040 ± 0.003, ICIR of 0.43 ± 0.03, RankIC of 0.048 ± 0.003, RankICIR of 0.50 ± 0.05, AR of 0.14 ± 0.05, and IR of 1.1 ± 0.3. The results show that the Informer model has a good performance in predicting the CSI300 and the BTC has some feasibility. The value of this paper is to provide investors of CSI300 and BTC with a feasible idea to assist trading.
Author: Yukun Niu
Institution: Xi’an Jiaotong-Liverpool University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
OpenAlex ID: W4405560103