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.
rainfull prediction
oussama_salhi89
Member Posts: 3 Learner III
hello
i'm a new user of RapidMiner, i will thinkfull if you help me to know how can i make a weather prediction (rainfull) from a database
Tagged:
0
Answers
Hi @oussama_salhi89,
It's difficult to help you without dataset. Can you share an extract of data of your database and explain what exactly you want to predict.
Regards,
Lionel
Hi @lionelderkrikor ;
thank you for replying me
this is my database
i have the mesure of rainfall since 1979 until 2011 in milimeter the object is to predict the evolution of rainfall in the comming years
/
Regards,
oussama
Hi again Oussama,
1. Here you can find a process for this time series project :
2. The parameters used in this process are (parameters you can set) :
- Window size = 10 years
- Horizon = 1 year
- Prediction Future Years = 7 years
3. I used a Deep Learning model, but you can try other models to improve the accuracy of your predictions.
4. This process contains pre-processing on the format of the date (for example 1979-1980 ==> 1979 , 1980-1981 ==> 1980, etc.).
5. You can find the Excel file I builded from you data by following this link :
https://drive.google.com/open?id=1hiq6L4zWI24E086Icq2jBcGkSHOARsup
I hope it helps,
Regards,
Lionel
hello again dear @lionelderkrikor
thank you very much to the solution
I know you gave me the solution but I am ashamed to say that I don't know how to apply it and I'am illiterate in RapidMiner I spend all the week to understand but I failed and I don't know how I can success my project graduation.
regards ;
oussama .
Hi @lionelderkrikor,
I try to import your XML Process to lean whats your solution, but with RM 8.1.000 (Free) and get a lot of warnings.
What is to do, to get it run inside RM 8.1.000 (Free)
Thanks!
Thomas
Hi @thomas_wiedmann,
you need to install value series extension. It's available in market place.
Best,
Martin
Dortmund, Germany
Hi @mschmitz,
thanks. I will checked this.
BTW I wanted to "click on like" for your solution, but it has no effect. Sorry.
Thanks!
Thomas
After get this extention from marketplace, everything work well.
A least one question: I have no idea why Date contain "Dec 28, yyyy" now in result. Why "Dec 28"?
Thanks !
Thomas
Hi Thomas,
Why "Dec 28"?
It's a good question and I will make a confession : I don't know why......
However good, in the nominal to date operator, I set "date format" as YYYY.
Or maybe in the Loop - > Increase date / Generates attributes operator when :
RapidMiner "forces" Year attribute to this format and value of "Dec 28"....
Best regards,
Lionel
@lionelderkrikor and @thomas_wiedmann I've encountered that issue before in a similar process and if I remember correctly it had to do with how the date-time was being intrepreted in the loop. I fixed it but I don't remember how I did it.
oh the Dec 28 is a RapidMiner easter egg - that's @IngoRM's birthday. :smileyvery-happy:
Sorry couldn't resist.
Scott
Hi @oussama_salhi89,
Here some elements to better understand the process :
At the beginning of the process, we define the following parameters :
- Window size : determine how many "attributes" are created for the cross -sectionnal data. Each row of the original time series
within the window width will become a new attribute. In our case, w = 10 (years)
- Horizon : Determine how far out to make the forecast. In our case, h = 1 (year).
Our window size = 10 and horizon = 1, so the 11th row of the original time series becomes the first sample for the "label" variable.
This figure shows the original data and the transformed output from the windowing process and describes the transformation details.
The main point is that for the window selected and shown in the box, the target value is the value of Jan 1, 1989.
When training, the model using this data, the attributes named Precipitations-9 through Precipitations-0 form the "independent variables".
Then, the trained model is applied to generate forecast.
I hope you understand better the "general philosophy" of this process.
Moreover I suggest you to set a Breakpoint After on each operator (via right-clicking on an operator), to better understand
the transformation of data step by step.
I hope it helps,
Regards,
Lionel
Hi,
@Thomas_Ott, @thomas_wiedmann
After trying several far fetched methods, I just add a Generate Attributes operator at the end of the process to convert
the date attribute to string. Now only the year is displayed.
Here the process :
Best regards,
Lionel
@lionelderkrikor
True, work fine..!
Thanks!
Thomas