Considering that the dataset has been cleaned and preprocessed, can a lack of correlation between the independent and dependent variables result in a negative $R^2$ score?

A negative $R^2$ score indicates that the slope does not fit the data trend properly. Could a lack of correlation within the dataset be a probable reason?

The algorithms used so far are DecisionTreeRegressor, RandomForestRegressor, and XGBRegressor. The default as well as modified parameters have been utilised. I'm trying to understand the factors that could be resulting in a negative $R^2$ score in this scenario. SelectKBest and chi2 have been used to determine high-ranking features and those features have been selected.

Would it be correct to say that high variation within the dataset can lead to a negative $R^2$ score?


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.