How I improved the accuracy of Kaggle’s May 2021 tabular competition by using SMOTE

For those individuals who have read my last post, they will know that I have been working on Kaggle’s May 2021 tabular competition. This competition is particularly problematic because the train file has a single column multiclass target, which must be converted to a one hot encoded target when the predictions are made. There is no clear direction on what the values in the submission should be, so this leaves the data scientist with the obligation to make his best guess.My most recent post on this subject can be…