Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.
Time Series data issue
Hi I'm new to Time Series ARIMA model. when I train this model with the data below. i got this error: Indices of the time series are not strictly monotonic increasing. I try to forcast energy usage for every 15 mins. please advise.
Time Usage
2/20/2018 1:00 PM | 1,327.68 |
2/20/2018 1:15 PM | 1,330.56 |
2/20/2018 1:30 PM | 1,351.68 |
2/20/2018 1:45 PM | 1,350.72 |
2/20/2018 10:00 AM | 1,302.72 |
2/20/2018 10:00 PM | 1,140.48 |
2/20/2018 10:15 AM | 1,305.60 |
2/20/2018 10:15 PM | 1,131.84 |
2/20/2018 10:30 AM | 1,305.60 |
Tagged:
0
Best Answer
-
tftemme Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 164 RM ResearchHi @Montse
Yes your data is not sorted correctly. You first have all years for January, then all years for February (so your 6 th. example is earlier than your 5 th. example). Hence the Arima Trainer correctly throws the error that your date indices are not montonic increasing. Please first sort your data after the date attribute using for example the Sort operator
Best regards,
Fabian
3
Answers
hello @xiangw02 welcome to the community! Some quick recommendations for you:
• Post your XML process here in this thread (see https://youtu.be/KkgB5QXWXJ8 and "Read Before Posting" on right when you reply)
• Attach your dataset if possible (use a fictionalized version if there are privacy concerns)
• Make sure you have all necessary extensions installed (see https://youtu.be/pjBqG3xtXx4)
Scott
Hi @xiangw02
Can you verify your process, your XML is broken...
Regards,
Lionel
Hi again @xiangw02,
OK, I can visualize your data in RapidMiner,
I understandand you want to forecast Energy (kW) in the future.....
but can you be more precise (in the last day, week, month......??????)
Regards,
Lionel
next 24 hours but by every 15 mins interval.
any one can help this?
Hi again @xiangw02,
I'm working on a process for your project.
I will have an answer element tomorrow.
Regards,
Lionel
@lionelderkrikor, did you figure out a solution? thanks!
Hi @xiangw02
I'm not a specialist of ARIMA, but you can find here a process to forecast the Energy using
the kNN model (you can test different models - SVM, Neural Nets etc. - inside the Validation operator in order to find the best performance).
I hope it helps,
Regards,
Lionel
Thanks. @lionelderkrikor. my data goes to 3/9/2018. the forcasting date is 3/18/2018. where do i set to 3/10/2018?
BTW can someone help debug my ARIMA process?
thanks
Hi again @xiangw02,
I think you have to preprocess your data, the time attribute (attribute Period) is not strictly increasing
and it generate a bug in the Arima trainer operator.
You have 2 times the same date-time in your dataset :
I hope it helps
Regards,
Lionel
@lionelderkrikor thanks again.
i reformat the date and Data Import wizard cannot format correct format. e.g. i attached the csv file for reference.
what's correct date-time format Data Import wizard can take?
Hi again @xiangw02,
This process work but i don't know why : the hours in the date_time attribute are not good displayed (always 12 : 00 :00 AM)
Regards,
Lionel
@lionelderkrikor, can you look at the solution you provided?
my data goes to 3/9/2018. the forcasting date is 3/18/2018. where do i set to 3/10/2018?
@lionelderkrikor, can you look at the solution you provided?
my data goes to 3/9/2018. the forcasting date is 3/18/2018. where do i set to 3/10/2018?
Hi @xiangw02,
In deed, this is a strange behaviour.
I will study that and will give you an answer element, i hope, soon .
Regards,
Lionel
Hi again @xiangw02,
I applied a "fix" to set the prediction to 3/10/2018.
Here the new process :
I hope it helps,
Regards,
Lionel
I have the same problem with a time series but I have not found any repeated date.
I attach you the file and next you can find the XML process, making only a classic decomposition of the data.
The error is the same " The indices of the time series are not strictly monotonic increasing. Check Indices Attribute: date"
I have checked out the "has indices parameter" and then works fine but I need the date associated with every row.
Thanks in advance.
Best regards,
Montse
Dortmund, Germany