python - ARIMA 模型抛出“LinAlgError（“SVD 未收敛”）”错误

Question

我的最终目标是预测给定服务器的 磁盘使用百分比（如 df -h 中的使用百分比：

$ df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/sda3              39G   31G  6.9G  82% /

)

将在未来使用机器学习。我正在利用 statsmodels python 库中的 ARIMA 模型进行时间序列预测。

我有一个数组，其中包含我记录了很多天的每日元组，格式为（epoch_timestamp，used_pct）：

results = [(1545346993, 80), (1545433403, 79), (1545519793, 80), (1545606202, 82), (1545692596, 80), (1545779002, 83), (1545865397, 77), (1545951799, 76), (1546038202, 75), (1546124601, 73), (1546210994, 73), (1546297394, 73), (1546383797, 73), (1546470197, 74), (1546556595, 74)]

如果您看到下面的最小代码示例，它会出错

“LinAlgError：SVD 没有收敛”。

有趣的是，如果你修改这个特定的元组 - results[6] 来自：

(1545865397, 77)

至：

(1545865397, 80)

然后代码按预期工作，输出如下：

predicted=77.635415, expected=73.000000
predicted=69.932872, expected=73.000000
predicted=76.074475, expected=73.000000
predicted=73.698213, expected=73.000000
predicted=72.721116, expected=74.000000
predicted=73.054932, expected=74.000000
Test MSE: 7.227

所以这告诉我某些特定于数据的原因导致了这种情况。通过研究其他问题，我确保没有 np.Nan 值或其他不良数据。无论如何，这是使用“坏”元组的完整示例：

import pandas as pd
import numpy as np
from sklearn.metrics import mean_squared_error
from statsmodels.tsa.arima_model import ARIMA


time_list = []
used_pct_list = []
results = [(1545346993, 80), (1545433403, 79), (1545519793, 80), (1545606202, 82), (1545692596, 80), (1545779002, 83), (1545865397, 77), (1545951799, 76), (1546038202, 75), (1546124601, 73), (1546210994, 73), (1546297394, 73), (1546383797, 73), (1546470197, 74), (1546556595, 74)]
for time, used_pct in results:
    time_list.append(time)
    used_pct_list.append(used_pct)

data = np.array(used_pct_list)
series = pd.Series(data,index=time_list)
X = series.values

size = int(len(X) * 0.66)
train, test = X[0:size], X[size:len(X)]
history = [x for x in train]

predictions = list()
for t in range(len(test)):
    model = ARIMA(history, order=(5,1,0))
    model_fit = model.fit(disp=0)
    output = model_fit.forecast()
    yhat = output[0]
    predictions.append(yhat)
    obs = test[t]
    history.append(obs)
    print('predicted=%f, expected=%f' % (yhat, obs))
error = mean_squared_error(test, predictions)
print('Test MSE: %.3f' % error)

以及它抛出的错误：

runfile('C:/Users/user/.spyder-py3/blarg_bad.py', wdir='C:/Users/user/.spyder-py3')
C:\Users\user\Anaconda3\lib\site-packages\scipy\signal\signaltools.py:1341: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
  out_full[ind] += zi
C:\Users\user\Anaconda3\lib\site-packages\scipy\signal\signaltools.py:1344: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
  out = out_full[ind]
C:\Users\user\Anaconda3\lib\site-packages\scipy\signal\signaltools.py:1350: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
  zf = out_full[ind]
C:\Users\user\Anaconda3\lib\site-packages\statsmodels\tsa\tsatools.py:607: RuntimeWarning: invalid value encountered in true_divide
....
  File "C:\Users\user\Anaconda3\lib\site-packages\numpy\linalg\linalg.py", line 1562, in svd
    u, s, vh = gufunc(a, signature=signature, extobj=extobj)

  File "C:\Users\user\Anaconda3\lib\site-packages\numpy\linalg\linalg.py", line 98, in _raise_linalgerror_svd_nonconvergence
    raise LinAlgError("SVD did not converge")

LinAlgError: SVD did not converge

我是机器学习的新手，一直在使用这个网站作为 arima 模型的指南：https ://machinelearningmastery.com/arima-for-time-series-forecasting-with-python/

标签： pythonmachine-learningscikit-learnstatsmodelsforecasting