david-cortes
commented
2 years ago That's a scikit-learn-specific parameter that the devs there introduced in order to make the model predictions conform to their idioms. If you look at the docs, what that parameter does is set a threshold on outlier scores based on quantiles from the scores that you get on the training data. That should be very easy to mimic.
Thank you for providing this interesting package. I'd like to use this package instead of scikit-learn, however, I'm not getting the same results with scikit-learn. I used the following configuration:
IsolationForest(max_samples=100, random_state=self.rng, bootstrap=False, warm_start=True, n_jobs=None, contamination=0.003, verbose=0)Contamination, I believe, is the most significant parameter for my aim. Could you please tell me how I may achieve the same results?