We have Z-score, IQR etc to identify outliers in data. This could be used to eliminate outliers even in labels. For e.g. if the target variable is a housing price, we could use inter-quartile ranges to isolate outliers.

I am looking at an algorithm or a programming practice to isolate outliers in target variable data for classification exercises. I feel this is pertinent because training data could have samples with incorrect classifications. Maybe the source of error was manual oversight or maybe a freak sequence of events which normally could not have occurred. So eliminating these rows should ideally enable the algorithm to perform better classifications. Any help is appreciated


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.