Predicting High-Frequency Stock Movement with Differential Transformer Neural Network

Predicting stock prices has long been the holy grail for providing guidance to investors. Extracting effective information from Limit Order Books (LOBs) is a key point in high-frequency trading based on stock-movement forecasting. LOBs offer many details, but at same time, they are very noisy. This paper proposes differential transformer neural network model, dubbed DTNN, predict movement according LOB data. The model utilizes temporal attention-augmented bilinear layer (TABL) and convolutional (TCN) denoise In addition, prediction module captures dependency between time series. A proposed incorporated into extract messy chaotic can identify fine distinction adjacent slices We evaluate several datasets. On open benchmark FI-2010, our outperforms other comparative state-of-the-art methods accuracy F1 score. experiments using actual data, also shows great forecasting capability generalization performance.