Skip to content

Latest commit

 

History

History
13 lines (9 loc) · 715 Bytes

File metadata and controls

13 lines (9 loc) · 715 Bytes

Feature-importance-methods-of-simulated-binary-black-holes

In our work we studied almost 2 million simulations of binary stellar systems, carried out with the rapid-synthesis binary population code MOBSE, searching for the most important features that determine the eventual evolution of a binary stellar system into a merging BBH, source of GWs.

The Goal of the project is to find what features have the highest impact on the evolution of a binary system into a Binary Black Hole. To do this we implemented four methods:

  • Weights of linear SVM
  • Medium Difference in Impurity for RFC
  • Permutation importance
  • Feature dropping

We also used SHAP values. We draw our conclusions in the project presentation.