Machine Learning Feature Selection with Correlation Matrix

Create a JavaScript module that performs advanced feature selection using correlation matrix analysis for predictive modeling. Develop an algorithm that calculates Pearson correlation coefficients between multiple numerical variables, identifies multicollinearity, and recommends optimal feature subsets for machine learning models. Implement visualization using D3.js to render an interactive correlation heatmap, with tooltips showing detailed statistical relationships. The solution should handle large datasets efficiently and provide recommendations for feature reduction.