In CMD, a temporary confidence estimation module shows that, while most of the Random Forest Regressor and Linear Regressor predictors achieve a 100% training score, their performance (measured in terms of f1 score) is only 20%. Perhaps first see if a restricted set of maps (e.g. sum/avg/diff of columns) apply would fix this overfitting issue.
In CMD, a temporary confidence estimation module shows that, while most of the Random Forest Regressor and Linear Regressor predictors achieve a 100% training score, their performance (measured in terms of f1 score) is only 20%. Perhaps first see if a restricted set of maps (e.g. sum/avg/diff of columns) apply would fix this overfitting issue.