Use seaborn to represent your heatmap easily and quickly identify the most important factors. Also look for independent variables that have high correlations - this is an indication of codependency and could be neutral to bad for your model. A predictor that has nothing to do with danger could be highly correlated with a predictor that does predict danger well, and if you include the bad predictor it will decrease prediction strength. Use this step as a chance to take out codependent factors and reintroduce stronger predictors if you feel the need to. Here are some others to consider.
alyoshenka
commented
1 year ago
Recommended Strategy 4
Use seaborn to represent your heatmap easily and quickly identify the most important factors. Also look for independent variables that have high correlations - this is an indication of codependency and could be neutral to bad for your model. A predictor that has nothing to do with danger could be highly correlated with a predictor that does predict danger well, and if you include the bad predictor it will decrease prediction strength. Use this step as a chance to take out codependent factors and reintroduce stronger predictors if you feel the need to. Here are some others to consider.
Then drop hindrance columns.
Depends on #69