time attribute in time serie

ThiruThiru Member Posts: 100 Guru
dear all, when I use  time series data, (enclosing the process)

1. My data has attribute 'Time" , in which each row represent  'seconds'. ( 1 sec, 2 sec) etc.  When I import this data,
the data type for this indices attribute  considered as 'time'.   However while viewing results after modelling, the data is shown as 12:00 AM IST in all the rows.  MY query is how to represent the results with actual timing in increments of seconds.

2.  while using 'process windows' operator,  if I consider window size as : 5,  the attributes are generated for every 5 rows.  However, how does the  'Horizon attribute "is estimated for this window size of 5?    Here in my enclosed process, the labelled atribute is Fault.   I need to have an understanding on this, as Im feeding the generated data as input to my  Machine learning model , lets say: kNN.?

thanks.

regds
thiru


Answers

  • tftemmetftemme Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 164 RM Research
    Hi @Thiru

    1. The results view only have a specific resolution which only shows the minutes. I think you can even specify in which format date time data is shown in the settings. The data behind it, still has millisecond resolution so you can still work with it and there is no problem in the execution of the data.
    If you cannot change the settings for the visualization, you can also generate a new attribute which is just the number of seconds since a reference date and use this for showing. 
    I think you need to put in something like:
    date_diff(<reference_date>,date_attribute,SECONDS)

    2. The horizon attribute are added automatically to the output of the Process windows operator. If their is only 1 horizon attribute (e.g. the horizon size is set to 1), the attribute is also set to the Label role.

    Hopes this helps
    Fabian
  • ThiruThiru Member Posts: 100 Guru
    hi @tftemme

    thanks for your reply.

    Reg 2nd point:  i would like to have more clarification.   If window size : 10+1 ,  & if I have labels as F1 for first  3 rows and  and F5 for next 4 rows , F8 for next two rows  etc.... How particular label is assigned automatically to every set of window, i if I have more than one label for the rows within that window size? 
    i.e. How does it choose among F1 or F5 or F8  for particular set of rows of a single window.

    thanks.
    regds
    thiru
  • tftemmetftemme Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, RMResearcher, Member Posts: 164 RM Research
    Hi @Thiru

    I am not sure if I correctly understand you. If you have the label information distributed over several attributes, this is not supported with RapidMiner (and I think impossible in general, cause RapidMiner would not know for which windows which Attribute contains the Label). 

    If you mean, how the Process Windows operator determines which value is taken for the label of a window: If you have a window size of 10 and a horizon size of 1, the first 10 rows are taken for the window, while the 11 (10+1) value of the horizon attribute is taken as the label.

    Please have also a look into the tutorial processes of the Process Windows operator (you could also check out the Windowing operator) to see the effect of windowing.

    Best regards,
    Fabian
Sign In or Register to comment.