statsmodels.discrete.discrete_model.NegativeBinomial.fit()提出LinAlgError：“奇异矩阵”

Question

我将statsmodels.discrete.discrete_model.NegativeBinomial用于负二项式回归任务，因此我使用以下脚本创建了一个模型：

from statsmodels.discrete.discrete_model import NegativeBinomial
#create a model
regr = NegativeBinomial(y_train, X_train)

在这里，我的y_train和X_train分别有<class 'numpy.ndarray'>类型和(276,) & (276, 252)形状。

我的问题是，当我调用regr.fit()时，会引发numpy.linalg.linalg.LinAlgError: Singular matrix错误。这是我的堆栈跟踪：

Traceback (most recent call last):
  File "/home/vajira/PycharmProjects/dengAI/neg_binomial_custom.py", line 137, in <module>
    regr_iq = regr_run(nptrain_iq, degree_iq, exploring=True)
  File "/home/vajira/PycharmProjects/dengAI/neg_binomial_custom.py", line 92, in regr_run
    regr.fit()
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/discrete/discrete_model.py", line 2756, in fit
    res_poi = mod_poi.fit(**optim_kwds_prelim)
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/discrete/discrete_model.py", line 1034, in fit
    disp=disp, callback=callback, **kwargs)
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/discrete/discrete_model.py", line 220, in fit
    disp=disp, callback=callback, **kwargs)
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/base/model.py", line 466, in fit
    full_output=full_output)
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/base/optimizer.py", line 191, in _fit
    hess=hessian)
  File "/home/vajira/ipython/lib/python3.6/site-packages/statsmodels/base/optimizer.py", line 278, in _fit_newton
    newparams = oldparams - np.dot(np.linalg.inv(H),
  File "/home/vajira/ipython/lib/python3.6/site-packages/numpy/linalg/linalg.py", line 528, in inv
    ainv = _umath_linalg.inv(a, signature=signature, extobj=extobj)
  File "/home/vajira/ipython/lib/python3.6/site-packages/numpy/linalg/linalg.py", line 89, in _raise_linalgerror_singular
    raise LinAlgError("Singular matrix")
numpy.linalg.linalg.LinAlgError: Singular matrix

有人能帮我解决这个问题吗？

Answer 1

我认为这是一个过于参数化的问题。看起来你有276个样本，有252个特征，这表明对于一个小样本来说，模型太复杂了。Singular matrix警告表明，该模型没有找到与该模型的最优收敛性。

我会回到过去，找出你对建模感兴趣的小得多的特性。