A Correlation-Embedded Attention Module to Mitigate Multicollinearity: An Algorithmic Trading Application

Algorithmic trading is a common topic researched in the neural network due to abundance of data available. It phenomenon where an approximately linear relationship exists between two or more independent variables. especially prevalent financial interrelated nature data. The existing feature selection methods are not efficient enough solving such problem potential loss essential and relevant information. These also able consider interaction features. Therefore, we proposed improvements apply Long Short-Term Memory (LSTM) this study. Multicollinearity Reduction Module (MRM) based on correlation-embedded attention mitigate multicollinearity without removing motivation allow model predict using relevance redundancy within first contribution paper allowing effects any second improving returns when our mechanisms applied LSTM. This study compared classification performance LSTM models with module. experimental result reveals that can learn improve desired performance. Furthermore, module 46.82% higher sacrificing training time. Moreover, MRM designed be standalone interoperable models.