The cross-interval price impact model and its empirical analysis on cryptocurrency order book

The demand for high-frequency algorithmic trading in the cryptocurrency markets is driving the research of price impact mechanisms. We propose the cross-interval price impact model (CIPIM) to explore the advanced or delayed price impact of order book events. The results of the empirical analysis show that neural network structures such as long short-term memory (LSTM) as a specific implementation of CIPIM obtain better concurrent interpretation on price impact than order flow imbalance (OFI) in Cont et al. ( J Financ Economet 12(1):47–88, 2014 ). Meanwhile, the classification version of CIPIM that predicts the direction of Bitcoin price changes tends to work to some extent.