Usage
def remove_outlier(df, col, partitionby=None, kernel='both', n_std=2, eps=1e-5):
"""
:param col: label column
:param partitionby: remove outliers in each partition
:param kernel: kernel options: ["std", "quartile", "both"]
:type df: pandas.DataFrame
:type col: str
:type partitionby: list
:type kernel: str
:type n_std: int
:type eps: float
:return: pandas.DataFrame
"""