CSVExampleSource and Numeric?

Stefan_E

Hi,
I try to load a .csv file with lines as so:
  64977,0.0542555,0.0531881,0.053273,0.0528606,0.0520237
through CSVExampleSource. The first column I declare as label. The rest gets converted to 'Nominal' instead of how I'd expect to Numeric.

I tried NominalNumbers2Numerical, but it does just nothing: the attributes stay nominal.

(I can work with ExampleSource - but what is CSVExampleSource good for then?)

Kind regards                      Stefan

Stefan_E

... same problem as in my other post on PCA: The dataset contained a line with all missing values except for the label. Once I remove this line, CSVExampleSource correctly recognizes the real values.

So, next time I'll print out my .csv files and read them in my leisure time before I give them to RM 

Stefan

martyns

Another one that I have found happening to me recently is that excel/spss often format numbers in their thousands with
1,234 for example.

This is a horrible thing to have happen when you save it as a csv!

steffen

Hello

Stefan_E wrote:
So, next time I'll print out my .csv files and read them in my leisure time before I give them to RM

Just a remark: You can use RM as an ETL tool, but this is not its primary focus. For this task other tools are more suitable. Beside... I tried to work with Pentaho Kettle and even this tool (meant as ETL) has problems to handle any kind of corrupted data.

martyns wrote:
Another one that I have found happening to me recently is that excel/spss often format numbers in their thousands with 1,234 for example.

I guess this is an adjustable option in excel

regards,

Steffen

land

Hi,
just a short comment: If you get your files into excel, you simply might import it from the excel file directly. This way you come around the problems with separator and number formating.

Greetings,
  Sebastian