Fig. 2.
Correlation matrix heat map showing the linear relationships between variables. ADL_related variables (e.g., Adl_standing, Adl_sitting, Adl_dressing) display strong positive correlations, suggesting redundancy. Features with weak correlations (e.g., Sex, Rehabilitation, Sedative_drugs) likely have limited impact on prediction. The strong correlations between Adl_sitting, Adl_dressing, and Adleating highlight potential multicollinearity, warranting the removal of redundant features
