PDA

View Full Version : SVM Data



huajie.lee
12-06-2007, 03:35 AM
Hi,

I would like to intend to use SVM algorithm using Weka on my input file.

However, I wonder what type of input file pattern is allowed. Currently, I am having columns with class label rather than the usual horizontal rows with class labels.

Do SVM allow these data?



Thanks.

-Jason

Mark
12-06-2007, 05:13 PM
Hi there,

Data will need to be in Weka's ARFF format, which is row-based (one row per training example). See here for details:

http://www.cs.waikato.ac.nz/~ml/weka/arff.html

Cheers,
Mark.

huajie.lee
12-06-2007, 06:57 PM
Hi Mark,

Thanks for your reply. I just wondered given the fact I have over 600 data attributes of numbers; with class label in the end, would I able to use SVM using Weka to produce some information?

I did try to use weather.arff and play around with SVM but I have difficulty in adapting my problem area in this.

For example;-

My scenario is having 1000 different groups which shares further 600 unique data attributes for each group respectively.

Example:-

Group Value
T1 2,5,8,9,3,.................(up to 600 number of attributes)
T2 .................

The group T1,T2..........(till 1000)

so I was wondering what would be my attribute look like? Would it be something like
@attribute TT{T1,T2.......T1000}
@attribute ?????
@attribute Get{Yes,No}

but how can I represent the vector of 600 numbers which each T1,T2......T1000 have?

Basically I would like to study the relationship of the 600 numbers which each element in TT have to predict the Get {Yes,No}

I am kind of new to Weka and so far the example I have seen is not on numeric but more on nominal such as weather.arff.



Please advise.


Thanks.

Rgrds,
Jason