netmanaiops / bagel Goto Github PK

Hi, author, there is an error after running main.py as shown in title, detail traceback as below:
Traceback (most recent call last):
File "c:/Users/admin/Desktop/study/Bagel-master/main.py", line 22, in
model.fit(train_kpi.label_sampling(0.), valid_kpi)
File "c:\Users\admin\Desktop\study\Bagel-master\model.py", line 132, in fit
drop_last=True)
File "c:\Users\admin\Desktop\study\Bagel-master\kpi_frame_dataloader.py", line 32, in init
self.dataset = dataset
File "C:\ProgramData\Anaconda3\envs\Bagel\lib\site-packages\torch\utils\data\dataloader.py", line 270, in setattr
'initialized'.format(attr, self.class.name))
I can't solve it, so call for help, thank u.

Hi, Zeyan Li. I have a question when applying the model on my datasets：

Bagel and donut assumes that the historical data follow normal pattern,
however, when the amount of historical data is not very large,
the impact of the abnormal points can not be ignored. So I want to ask that:
how to mitigate the impact of abnormal points in historical data on training?

I have tried to introduce the labels of the data into the model training, but it has not improved much.

I will be appreciated if you can help me to solve this problem, thank you.

请教下作者，输入的数据集是否有格式的要求？我使用了自己的数据集，数据集中的value值为0-100的数，timestamp为一天的时间范围，经过您的模型训练后的得到的f1score为0。

In main.py, we can only get the probability list, how can we transfer it to label list?

Hello, I have met some problems during my training, that is as long as I changed the parameters of train_kpi.label_sampling(0.) and test_kpi.label_sampling(0.) to (1.), the best F1- score became very low, which is less than 0.1. So I would appreciate it if you could tell me why the program samples some labels before traning and how will it affect the algorithm if I change the sampling ratio. Thank you!

Hi guys,

Thanks for creating this package. I see this package works good on datasets with no missing values but one of the dataset I was testing had missing timestamps. My data is KPI data i.e Sales every 5 minute in a store.

Your KPI class has a property called missing value which is causing my raw_data to be negative. For example, if a data point of 19-June-2020 3:35:00 pm is missing then this value is getting replaced by some negative number. And the negative number is the result of this function
@Property
def missing_value(self):
return self.value[self.missing == 1][0] if np.count_nonzero(self.missing) > 0 else 2 * np.min(self.value) -np.max(self.value)

this property is causing my raw data to be negative when in reality my sales can never be negative.

1. Could you tell me what's the purpose of this function?
2. In order to fix this issue, should I pass an array when calling KPISeries class to be assigned to missing parameter(which by default is None)?

• If yes, what does that array contain? Should it be values or timestamps of missing points?

Let me know how you've designed it. I am sure I am missing something here
Appreciate your help!

Hi,

I can currently run your code in a VS code environment and can see the network being trained but I don't see any visualizations. Is there something I need to do to see them? thanks.

版本冲突 py37下mltoolkit 版本错误

Hi Zeyan Li,
It is great to see your codes in Git Hub. I have downloaded them using git clone. I am very much interested in deep generative learning models. It looks like your work is one of the latest in RNN-VAE domain. I want to execute Bagel on pendigit dataset (http://odds.cs.stonybrook.edu/pendigits-dataset/) . Then I want to implement my own proposition on top of Bagle.
I will cite your paper in my research paper. It will be a great help if you upload a python code to execute Bagle for any Univariate/Multivariate Timeseries dataset.

在threshold_ml里面有一个threshold_prior，我没有找到threshold_prior这个函数，请问这个threshold_prior是哪个？

老铁，我觉得你的detect好像得叫predict啊