Converting the non numerical features to numeric representation would have been a great opportunity to use one hot encoding. In short, by using a scale instead of One-hot Encoding the student may have inadvertently given some features(E.g. Age) imbalanced significance (I.e. 45 could be considered more significant than 15 solely based on how large the number is not the actually feature itself). I'm certain if the student had used this technique, the models accuracy would have improved.
Converting the non numerical features to numeric representation would have been a great opportunity to use one hot encoding. In short, by using a scale instead of One-hot Encoding the student may have inadvertently given some features(E.g. Age) imbalanced significance (I.e. 45 could be considered more significant than 15 solely based on how large the number is not the actually feature itself). I'm certain if the student had used this technique, the models accuracy would have improved.
Here's a link if the student is interested learning more: https://machinelearningmastery.com/why-one-hot-encode-data-in-machine-learning/