This study develops a novel deep learning framework for bank distress prediction, designed to overcome the limitations of static network analysis and to enhance model interpretability. We propose a Spatio-Temporal Attention Network that uniquely captures the time-varying nature of systemic risk. Methodologically, it introduces two key innovations: (1) a dynamic interbank network whose connection weights are adjusted by the volatility of the Shanghai Interbank Offered Rate (SHIBOR), reflecting real-time market liquidity changes; and (2) a dual spatio-temporal attention mechanism that identifies critical time steps and pivotal contagion pathways leading to a distress event. Empirical results demonstrate that the model significantly outperforms traditional benchmarks across key metrics including accuracy and F1-score. Most critically, the architecture proves exceptionally effective at reducing Type II errors, substantially minimizing the failure to identify at-risk banks. The model also offers high interpretability, with attention weights visualizing intuitive risk evolution patterns. We conclude that incorporating dynamic, liquidity-adjusted networks is crucial for superior predictive performance in systemic risk modeling.

Deep learning technology is rapidly adopted in financial market settings. Using a large data set from the Chinese stock market, we propose a return-risk trade-off strategy via a new transformer model. The empirical findings show that these updates, such as the self-attention mechanism in technology, can improve the use of time-series information related to returns and volatility, increase predictability, and capture more economic gains than other nonlinear models, such as LSTM. Our model employs Shapley additive explanations (SHAP) to measure the “economic feature importance” and tabulates the different important features in the prediction process. Finally, we document several economic explanations for the TF model. This paper sheds light on the burgeoning field on asset allocation in the age of big data.

Stock market prediction provides the decision-making ability to the different stockholders for their investments. Recently, stock technical indicators (STI) emerged as a vital analysis tool for predicting high-dimensional stock data in various studies. However, the prediction performance and error rate still face limitations due to the lack of correlational analysis between STI and stock movement. This paper proposes a correlational strategy to overcome these challenges by analyzing the correlation of STI with stock movement using neural networks with the feature vector. This strategy adopts the Pearson coefficient to analyze STI and close index of stock data from 8 Chinese companies in the Hong Kong stock market. The results reveal the price prediction of BiLSTM outperformed the GRU and LSTM in various datasets and prior studies.

This paper uses ensemble learning to build a predictive model to analyze the short selling mechanism of short institutions. We demonstrate the value of combining domain knowledge and machine learning methods in financial market. On the basis of the benchmark model, we use three input data: stock price, financial data and textual data and we employ one of the most powerful machine learning methods, ensemble learning, rather than the commonly used method of logistic regression. In specific methods, we use LSTM-AdaBoost and CART-AdaBoost for model prediction. The results show that the model we train have strong prediction ability for short-selling and the company' s financial text data is more likely to have an impression of whether it would be shorted or not.