1
$\begingroup$

I am working on a classification model using one of the following three algorithms: RandomForestClassifier, a TensorFlow model and a LogisticRegression model.

The data set I am working with has a feature that is represented by a single word that uses ASCII characters (may or may not be a valid word in any language). I don't see any advantage in treating this column as categorical data since number of unique words/total number of rows is very close to 1, i.e., almost every word is unique.

Is there any obvious way to use this column to improve the predictive capabilities of the resulting classification model?

The data I am working with are player IDs that are strings and are meaningless in any language. But would the answer to the above question change if I were working with single English words?

$\endgroup$

1 Answer 1

1
$\begingroup$

In general there is no reason to include a meaningless id as feature, it has no semantic value.

At the semantic level, the question is whether knowing this information provides any help with knowing the target label. In other words, would a human expert be able to use this information? If not it's very unlikely to be relevant.

At the technical level, you can measure the amount of information brought by this variable about the target, for example with conditional entropy.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.