Given some number of features and target, how can i show that features are independent of target? I found that without any features my model perform well, but with them score is very poor. My hypothesis is that features have no impact (i.e their combination) on target, but is there a precise way to prove this? Maybe some manifold learning methods?
Asked
Active
Viewed 46 times
0
-
Are you using accuracy with an imbalanced dataset? Be careful so the model does not always predict the same value, e.g., always predicting 1 when 1 is the majority class. – Shaido Mar 09 '21 at 09:24
-
@Shaido no, i am working with time series with smape score. So i want to check whether some macro-parameters have impact on target – friendlyuser Mar 09 '21 at 09:59