按类型和以前的计数预测错误
我有以下时间序列数据集:
time error count
0 2021-10-03 1111 2
1 2021-10-03 2222 4
2 2021-10-03 3333 6
3 2021-10-03 4444 2
4 2021-10-03 5555 8
我创建了一个带有属性“时间”,“ error”和“ count”的类error_type,然后将这些类插入词典,其中关键为时间(出现日期),值是具有上述对象属性。
因此,字典看起来像:
time (2021-10-03)
- class with error (1111) and count (2)
- class with error (2222) and count (4)
.
.
.
time (2021-10-04)
- class with error (3333) and count (6)
- class with error (4444) and count (2)
.
.
.
我有几个月的数据,它们看起来如下:
我想完成的是仔细阅读字典,并通过错误类型预测第二天的错误计数,因此我们知道我们可以期望下几天有多少错误以及我们可以期望的类型。
我与这里的时间序列相混淆,并试图将Arima模型仅用于预测错误计数的预测,而错误计数无法正常工作。我试图更改周末,在滚动率较低的情况下,删除了没有发生错误的日子,但这并没有帮助Arima进行准确的预测。
I have the following time-series dataset:
time error count
0 2021-10-03 1111 2
1 2021-10-03 2222 4
2 2021-10-03 3333 6
3 2021-10-03 4444 2
4 2021-10-03 5555 8
I created a class error_type with attributes 'time', 'error' and 'count' and then insert these classes to dictionary where key is time (date of occurrence) and value is the object with mentioned attributes.
So the dictionary looks like:
time (2021-10-03)
- class with error (1111) and count (2)
- class with error (2222) and count (4)
.
.
.
time (2021-10-04)
- class with error (3333) and count (6)
- class with error (4444) and count (2)
.
.
.
I have data for several months and they look as follows:
What I would like to accomplish is go through dictionary and predict count of errors in total for the next day by error type so we know how many errors we can expect next days and what type we could expect.
I am confused with the time-series here and tried to apply ARIMA model just for prediction of error counts which did not work well. I tried to change weekends where there are lower numbers to rolling mean, delete those days where no error happened but it did not help ARIMA to give accurate prediction.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论