使用 sklearn 训练不同的回归器

Question

我有一个Xs 的列表及其输出值Ys。并使用以下代码，我可以训练以下回归器：

• 线性回归器
• 等渗回归器
• 贝叶斯岭回归器
• 梯度提升回归器

代码：

import numpy as np

from sklearn.linear_model import LinearRegression, BayesianRidge
from sklearn.isotonic import IsotonicRegression
from sklearn import ensemble
from sklearn.svm import SVR
from sklearn.gaussian_process import GaussianProcess


import matplotlib.pyplot as plt
from matplotlib.collections import LineCollection


def get_meteor_scores(infile):
    with io.open(infile, 'r') as fin:
        meteor_scores = [float(i.strip().split()[-1]) for 
                               i in re.findall(r'Segment [0-9].* score\:.*\n', 
                                               fin.read())]
        return meteor_scores

def get_sts_scores(infile):
    with io.open(infile, 'r') as fin:
        sts_scores = [float(i) for i in fin]
        return sts_scores

Xs = 'meteor.output.train'
Ys = 'score.train'
# Gets scores from https://raw.githubusercontent.com/alvations/USAAR-SemEval-2015/master/task02-USAAR-SHEFFIELD/x.meteor.train
meteor_scores = np.array(get_meteor_scores(Xs))
# Gets scores from https://raw.githubusercontent.com/alvations/USAAR-SemEval-2015/master/task02-USAAR-SHEFFIELD/score.train
sts_scores = np.array(get_sts_scores(Ys))

x = meteor_scores
y = sts_scores
n = len(sts_scores)

# Linear Regression
lr = LinearRegression()
lr.fit(x[:, np.newaxis], y)

# Baysian Ridge Regression
br = BayesianRidge(compute_score=True)
br.fit(x[:, np.newaxis], y)

# Isotonic Regression
ir = IsotonicRegression()
y_ = ir.fit_transform(x, y)

# Gradient Boosting Regression
params = {'n_estimators': 500, 'max_depth': 4, 'min_samples_split': 1,
          'learning_rate': 0.01, 'loss': 'ls'}
gbr = ensemble.GradientBoostingRegressor(**params)
gbr.fit(x[:, np.newaxis], y)

但是如何为Support Vector Regression、Gaussian Process 和Decision Tree Regressor 训练回归器？

---

当我尝试以下方法训练 Support Vector Regressors 时，出现错误：

from sklearn.svm import SVR
# Support Vector Regressions
svr_rbf = SVR(kernel='rbf', C=1e3, gamma=0.1)
svr_lin = SVR(kernel='linear', C=1e3)
svr_poly = SVR(kernel='poly', C=1e3, degree=2)
y_rbf = svr_rbf.fit(x, y)
y_lin = svr_lin.fit(x, y)
y_poly = svr_poly.fit(x, y)

[出]：

Traceback (most recent call last):
  File "/home/alvas/git/USAAR-SemEval-2015/task02-somethingLiddat/carolling.py", line 47, in <module>
    y_rbf = svr_rbf.fit(x, y)
  File "/home/alvas/.local/lib/python2.7/site-packages/sklearn/svm/base.py", line 149, in fit
    (X.shape[0], y.shape[0]))
ValueError: X and y have incompatible shapes.
X has 1 samples, but y has 10597.

当我尝试Gaussian Process时发生同样的情况：

from sklearn.gaussian_process import GaussianProcess
# Gaussian Process
gp = GaussianProcess(corr='squared_exponential', theta0=1e-1,
                     thetaL=1e-3, thetaU=1,
                     random_start=100)
gp.fit(x, y)

[出]：

Traceback (most recent call last):
  File "/home/alvas/git/USAAR-SemEval-2015/task02-somethingLiddat/carolling.py", line 57, in <module>
    gp.fit(x, y)
  File "/home/alvas/.local/lib/python2.7/site-packages/sklearn/gaussian_process/gaussian_process.py", line 271, in fit
    X, y = check_arrays(X, y)
  File "/home/alvas/.local/lib/python2.7/site-packages/sklearn/utils/validation.py", line 254, in check_arrays
    % (size, n_samples))
ValueError: Found array with dim 10597. Expected 1

运行gp.fit(x[:,np.newaxis], y) 时出现此错误：

Traceback (most recent call last):
  File "/home/alvas/git/USAAR-SemEval-2015/task02-somethingLiddat/carolling.py", line 95, in <module>
    gp.fit(x[:,np.newaxis], y) 
  File "/home/alvas/.local/lib/python2.7/site-packages/sklearn/gaussian_process/gaussian_process.py", line 301, in fit
    raise Exception("Multiple input features cannot have the same"
Exception: Multiple input features cannot have the same target value.

当我尝试Decision Tree Regressor:

from sklearn.tree import DecisionTreeRegressor
# Decision Tree Regression
dtr2 = DecisionTreeRegressor(max_depth=2)
dtr5 = DecisionTreeRegressor(max_depth=2)
dtr2.fit(x,y)
dtr5.fit(x,y)

[出]：

Traceback (most recent call last):
  File "/home/alvas/git/USAAR-SemEval-2015/task02-somethingLiddat/carolling.py", line 47, in <module>
    dtr2.fit(x,y)
  File "/home/alvas/.local/lib/python2.7/site-packages/sklearn/tree/tree.py", line 140, in fit
    n_samples, self.n_features_ = X.shape
ValueError: need more than 1 value to unpack

Answer 1

所有这些回归器都需要 多维 x 数组，但您的 x 数组是一维数组。因此，唯一的要求是将 x-array 转换为 2D 数组以使这些回归器起作用。这可以使用x[:, np.newaxis]来实现

演示：

>>> from sklearn.svm import SVR
>>> # Support Vector Regressions
... svr_rbf = SVR(kernel='rbf', C=1e3, gamma=0.1)
>>> svr_lin = SVR(kernel='linear', C=1e3)
>>> svr_poly = SVR(kernel='poly', C=1e3, degree=2)
>>> x=np.arange(10)
>>> y=np.arange(10)
>>> y_rbf = svr_rbf.fit(x[:,np.newaxis], y)  
>>> y_lin = svr_lin.fit(x[:,np.newaxis], y)
>>> svr_poly = svr_poly.fit(x[:,np.newaxis], y)
>>> from sklearn.gaussian_process import GaussianProcess
>>> # Gaussian Process
... gp = GaussianProcess(corr='squared_exponential', theta0=1e-1,
...                      thetaL=1e-3, thetaU=1,
...                      random_start=100)
>>> gp.fit(x[:, np.newaxis], y)
GaussianProcess(beta0=None,
        corr=<function squared_exponential at 0x7f46f3ebcf50>,
        normalize=True, nugget=array(2.220446049250313e-15),
        optimizer='fmin_cobyla', random_start=100,
        random_state=<mtrand.RandomState object at 0x7f4702d97150>,
        regr=<function constant at 0x7f46f3ebc8c0>, storage_mode='full',
        theta0=array([[ 0.1]]), thetaL=array([[ 0.001]]),
        thetaU=array([[1]]), verbose=False)
>>> from sklearn.tree import DecisionTreeRegressor
>>> # Decision Tree Regression
... dtr2 = DecisionTreeRegressor(max_depth=2)
>>> dtr5 = DecisionTreeRegressor(max_depth=2)
>>> dtr2.fit(x[:,np.newaxis],y)
DecisionTreeRegressor(compute_importances=None, criterion='mse', max_depth=2,
           max_features=None, min_density=None, min_samples_leaf=1,
           min_samples_split=2, random_state=None, splitter='best')
>>> dtr5.fit(x[:,np.newaxis],y)
DecisionTreeRegressor(compute_importances=None, criterion='mse', max_depth=2,
           max_features=None, min_density=None, min_samples_leaf=1,
           min_samples_split=2, random_state=None, splitter='best')

GaussianProcess 的预处理：

xu = np.unique(x)  # get unique x values
idx = [np.where(x==x1)[0][0] for x1 in xu]  # get corresponding indices for unique x values
gp.fit(xu[:,np.newaxis], y[idx])  # y[idx] selects y values corresponding to unique x values

Answer 2

Multiple input features cannot have the same target value.

这意味着一个数据点在您的输入数据中重复，并且高斯过程不允许一个数据点被列出两次。 不幸的是，您的数据集不再可用，所以我无法检查这个，但我认为应该是这样。