def RemoveMultiOutliers(dataset_X, dataset_Y, col1, col2):
# fit the model for outlier detection (default)
clf = LocalOutlierFactor(n_neighbors=20, contamination=0.01)
# use fit_predict to compute the predicted labels of the training samples
# (when LOF is used for outlier detection, the estimator has no predict,
# decision_function and score_samples methods).
X = list(
zip(dataset_X[col1].values.tolist(), dataset_X[col2].values.tolist()))
y_pred = clf._fit_predict(X)
to_remove = set()
for i in range(len(X)):
if y_pred[i] == -1:
to_remove = to_remove | (set(
dataset_X.loc[(dataset_X[col1] == X[i][0])
& (dataset_X[col2] <= X[i][1])].index.tolist()))
to_remove = list(to_remove)
dataset_X.drop(to_remove, inplace=True)
dataset_Y.drop(to_remove, inplace=True)
