Question

我想知道为什么即使在使用ARIMA(1,1,1)时将微分阶数指定为1，也仍然会得到“计算出的初始AR系数不是固定的” 。 df['prod rate']包含从1到20的升序值，我很确定要删除趋势，我需要在此处应用一阶差分。

我已经通过以下答案链接：

但是我找不到解决问题的办法。

from statsmodels.tsa.arima_model import ARIMA
plt.figure(figsize = (10,6))
model = ARIMA(df['prod rate'], order = (1,1,1))  
results_AR = model.fit()
plt.plot(df['prod rate'], label = "Original")
plt.plot(results_AR.fittedvalues, color = 'red', label = 'Predictions')
plt.legend(loc = 'best')

在拟合模型时出现以下错误：

ValueError                                Traceback (most recent call last)
<ipython-input-59-2b92bbf4a4de> in <module>()
      2 plt.figure(figsize = (10,6))
      3 model = ARIMA(df['prod rate'], order = (1,1,1))
----> 4 results_AR = model.fit()
      5 plt.plot(df['prod rate'], label = "Original")
      6 plt.plot(results_AR.fittedvalues, color = 'red', label = 'Predictions')

3 frames
/usr/local/lib/python3.6/dist-packages/statsmodels/tsa/arima_model.py in _fit_start_params_hr(self, order, start_ar_lags)
    539         if p and not np.all(np.abs(np.roots(np.r_[1, -start_params[k:k + p]]
    540                                             )) < 1):
--> 541             raise ValueError("The computed initial AR coefficients are not "
    542                              "stationary\nYou should induce stationarity, "
    543                              "choose a different model order, or you can\n"

ValueError: The computed initial AR coefficients are not stationary
You should induce stationarity, choose a different model order, or you can
pass your own start_params.

为什么即使在对我的时间序列进行了差分处理后，为什么仍然得到“计算出的初始AR系数不是固定的”？

0 个答案: