Sensitivity analysis based on variance

Question

Please help me in understanding the sensitivity analysis based on variance in ml models

I am working on a binary classification problem apart from predicting the target variable I am also interested in finding the most important features contributing to the occurrence of specific class.

To achieve this I'm performing sensitivity analysis;I am using k cross validation and I am removing one column at a time and calculating the change in variance when all the columns are included

I expected the variable which when removed has got the highest variance, to be present at top of random forest and decision tree but it's not at the top

Please help me understand how the sensitivity analysis I'm doing is different from the variables at the top of random forest

Thank you

Answer 1

Sensitivity analysis based on variance assesses how uncertainty in input parameters impacts a model's output variability. It involves:

1. Identifying key input parameters.
2. Defining ranges or distributions for these parameters.
3. Running simulations or calculations with varied inputs.
4. Analyzing how input variations affect output variability.
5. Prioritizing influential parameters.
6. Using insights for decision-making or risk assessment.

This method quantifies the impact of uncertainties, helping prioritize factors affecting a model's results.