Forecasting Stock Price Volatility Based on Lstm-Transfomer - A Case of Pharmaceutical and Biological Industry

This paper establishes a featured indicator system comprising objective and subjective factors. The objective factors encompass basic and technical indicators of stocks within the pharmaceutical industry, macroeconomic factors, and industry-specific information. Principal components containing industry information and macroeconomic data are extracted using principal component analysis. Subsequently, the featured indicator system is modeled using an LSTM-Transformer composite model. The Long Short-Term Memory (LSTM) model effectively captures the characteristics of stock time series data, while the Transformer, with its self-attention mechanism, addresses the issue of long-term dependence. The model is evaluated using a dataset consisting of four pharmaceutical industry stocks, yielding favorable results.