Datetime error

[Marlina]

Hi all,
 I'm currently working on Holt's Winter modeling and encountering this error. I need your advice on how to solve this.

Error in eval(script, envir = .GlobalEnv) : Please include a time column of date/datetime type from the training dataset.
eval(script, envir = .GlobalEnv)
eval(script, envir = .GlobalEnv)
withCallingHandlers({
stop("Please include a time column of date/datetime type from the training dataset.")

[David Boot-Olazabal]

Hi Marlina,

Can you elaborate a bit more on what you have done? Is this in Spotfire or is this in Statistica? And can you share your dataset (or paste a screenshot) so we can see if you have included a time column of the requested date/datetime format?

Kind regards,

David

[Marlina]

Hi David, Thank u so much for responding my post. Pls find attachment to explain my data table and steps. Appreciate your advise. I have been working on this for weeks.

 

I refer to spotfire utube as per link 

 

 

Need help from spotfire team.pptx

[Gaia Paolini]

From your setting, Time Column does contain a date but its data type is shown as string. You need to change the data type when you import the dataset.
 

[Gaia Paolini]

Note that the Dr Spotfire video refers to the previous version of this data function

[Marlina]

Hi Gaia,

Can u share which tutorial video spotfire latest version of Holt data function

[Marlina]

Hi Gaia,

If u read the error, it refers to the data in training dataset, not in timecolumn. Training dataset i already choose date datatype.

\

 

 

When I read through read me pdf from spotfire, 

 

It mention that the time_column data type allowed is string as i marked in yellow

[Gaia Paolini]

time_column is the name of the column, not the column itself. You see is not a column but a value.
There is no training video for the new version of Holt Winters. There is an example Spotfire dxp that is downloaded with the data function:

 

[Marlina]

Hi Spotfire team, 

I've solved the date time error, but i encountered new error. Pls find my dxp and step by step how i use Holts Modelling

Need help from spotfire team.pptx OQA and In-Process Prediction 140624-for review.dxp

[Gaia Paolini]

Your data is not ready to be used in the Holt Winters application:
1) there are multiple values of cpk  per date. Your cpk column has up to 1340 values for the same date. Is this right or do you need to slice by some other column? The supplied pre-processing data function can aggregate multiple values, as long as you think this is correct. Note that the Holt Winters application provided does not do forecast by group.
2) the data is not a regular time series. There are gaps, sometimes very big in the data. You can close some gaps by aggregating via the pre-processing data function.
I have tested it with the pre-processing and after aggregation the data appears weekly and has only 77 values.

However, I am seeing an issue with the TERR engine (I am on Spotfire 14.4). If you are on an earlier Spotfire version can you do the following test?
Upload your dataset as the Training dataset into the provided example Spotfire dxp. Use the pre-processing data function before the Holt Winters one. Set the Test data to None.

 

 

[Gaia Paolini]

I tried on Spotfire 14.3 and it works. However, your data does not have any seasonal trend, so I suspect Holt Winters is not the appropriate method:
setting the tolerance very high during pre-processing, I get a lot of interpolated data on a daily basis.
applying Holt Winters gets me a forecast but only it I set the frequency manually, and I do not see any frequency in the data

[Gaia Paolini]

The way to make it work with automatic frequency is to set Gamma=false. But that takes away the seasonality, so the advantage of Holt Winters.

[Marlina]

Hi Gaia, thank u so much for your advise. Yes, my data has no seasonality and i also plan to set gamma as 0 to remove the seasonality. My study on Holt Winter article which mention about pre-processing data function only available for spotfire 12.5. As I'm using the below version Spotfire, pls advise how can i pre-process my data.

 

[Marlina]

Hi Gaia,

Regarding your question on why multiple CPK per date.

The data submitted by different suppliers, by different program, and 1 program could have multiple critical parameters. Each critical parameter will have its own CPK (supplier will measure the dimension). That is why it seems to have multiple CPK per date but it from different supplier, program and critical parameter.

[Gaia Paolini]

Your version should be ok. Please download the example dxp that comes with the Holt Winters data functions in the Community. You will find that there is already a pre-processing data function.
The only thing you need to make sure of is to set the Test data input parameter to None in both the pre-processing and the Holt Winters data functions.

In terms of data, it is your business case that will dictate how you aggregate it. I don't know the meaning of the different columns, I can only advise to look into it. It seems that at any given date, the number of cpk values varies from 1 to 1300, so not all the suppliers, programs and critical parameters are represented in all dates.