machine-learning - Pipeline 的最后一步应该实现 fit 或者是字符串 'passthrough'

Question

我正在尝试构建简单的管道:

from sklearn.linear_model import Lasso
from sklearn.preprocessing import PolynomialFeatures
from sklearn.pipeline import make_pipeline


make_pipeline([
    ('PolynomialFeatures', PolynomialFeatures(include_bias=False)),
    ('Lasso', Lasso(fit_intercept=True, max_iter=1000))])

我收到错误:

TypeError: Last step of Pipeline should implement fit or be the string 'passthrough'. '[('PolynomialFeatures', PolynomialFeatures(include_bias=False)), ('Lasso', Lasso())]' (type <class 'list'>) doesn't

怎么了？我该如何解决？

Answer 1

您应该使用 pipeline 而不是 make_pipeline 因为您提供了带名称的步骤(更多信息 here !)。

from sklearn.linear_model import Lasso
from sklearn.preprocessing import PolynomialFeatures
from sklearn.pipeline import make_pipeline, Pipeline


Pipeline([
    ('PolynomialFeatures', PolynomialFeatures(include_bias=False)),
    ('Lasso', Lasso(fit_intercept=True, max_iter=1000))])

输出:

Pipeline(steps=[('PolynomialFeatures', PolynomialFeatures(include_bias=False)),
                ('Lasso', Lasso())])