Stock price prediction using long short-term memory neural networks

PCA results show that the first few principal components account for over 95% of the total variance, enabling effective dimensionality reduction and extraction of core information. Outlier detection via box plots reveals abnormal fluctuations in indicators for companies like CATL and CRRC, providing clues to firm-specific risks. Pearson correlation analysis highlights strong linear relationships between indicators—such as a 0.96 correlation between CP and GPM—offering insights into potential co-movement and decision relevance. The proposed framework not only facilitates comprehensive financial data interpretation but also supports market risk assessment, indicator linkage analysis, and investment strategy optimization, thereby expanding the practical scope of multi-indicator analysis in the financial domain.