Question

我编写了一个自定义的Transformer对象，它引入了一个时间延迟的信号嵌入。我想将其超参数包含到GridSearchCV中。问题是它需要一个后处理步骤来删除从pandas shift操作引入的空值的行。据我了解，变形金刚不是为了处理目标而设计的，尽管我找到了一种方法来实现目标。问题是，这是一种合法的方法，还是有一种不太常见的方法呢？

时间延迟嵌入变换器

from pandas import Series, DataFrame
from pandas.tools.merge import concat
from sklearn.base import TransformerMixin
from typing import Union


class TimeDelayEmbedder(TransformerMixin):
    """
    A Transformer to add time-delay embedding to an existing time history.
    """
    def __init__(self, min_delay: int, max_delay: int):
        """
        Create a new TimeDelayEmbedder.

        :param min_delay: The minimum delay to use.
        :param max_delay: The maximum delay to use.
        """
        self._min_delay = min(min_delay, max_delay)
        self._max_delay = max(min_delay, max_delay)

    def fit(self, X=None, y=None):

        return self

    def transform(self, X: Union[Series, DataFrame], y: Series=None):
        """
        Transform the input data `X`. The returned DataFrame will have null values,
        which will need to be handled separately e.g. with a NullRowDropper.

        :param X: A pandas Series or DataFrame of time histories.
            For DataFrames, each column will have a history added.
        :param y: The targets. These will be unaffected.
        :rtype: DataFrame
        """
        if type(X) is Series:
            assert X.name, 'Input Series must have a name.'
            X = X.to_frame(name=X.name)
        series_list = []
        for column in X.columns:
            for time_step in range(self._min_delay, self._max_delay + 1):
                data = X[column].shift(time_step)
                if time_step < 0:
                    name = '%s__t+%i' % (column, time_step)
                elif time_step == 0:
                    name = '%s__t' % column
                else:
                    name = '%s__t-%i' % (column, time_step)
                series_list.append(Series(data=data, name=name))
        df_out = concat(series_list, axis=1)

        return df_out

空行丢弃转换器是：

from pandas import DataFrame, Series
from sklearn.base import TransformerMixin


class NullRowDropper(TransformerMixin):
    """
    Drop null rows from a data-set.
    Basically a pass-through to pandas dropna() but drops associated rows on the targets too.
    """
    def __init__(self, how='any'):
        """
        Create a new NullRowDropper.

        :param how: `'any'` or `'all'`
        """
        assert how in ('any', 'all'), "param: 'how' must be either 'any' or 'all'"
        self._how = how

    def fit(self, X: DataFrame=None, y: Series=None):

        self._y = y  # save y here so can modify its index inplace later (y is not passed to transform)
        return self

    def transform(self, X: DataFrame, y: Series=None):
        """
        Drop the null rows according to the `how` param given at initialisation.
        :param X: The input pandas DataFrame.
        :rtype: DataFrame
        """
        X_out = X.dropna(how=self._how)
        self._y.drop(set(self._y.index) - set(X_out.index), inplace=True)
        return X.dropna(how=self._how)

使用scikit-learn管道中的输入转换目标

0 个答案: