Activity 3.1 Question 2 - petalwidth or petallength?

[ngk]

Minor question.. When I run OneR (10-fold cross-validation) on the iris dataset I get the following model for the full dataset:

petallength:

< 2.45 -> Iris-setosa

< 4.85 -> Iris-versicolor

>= 4.85 -> Iris-virginica

(143/150 instances correct)

The Activity says the correct answer should be "petalwidth" rather than "petallength". Am I missing something?

[Thales Maciel]

Hello, Kelly.

I tried out some different options for testing, but wasn't able to figure out what, exactly, is that you set that provided you the result you mentioned. Petalwidth is correct, by the way (92% in accuracy). =)

Below is what you can do to make sure you are not missing anything.

1. Make sure you are using the latest stable version of WEKA, which is 3.6.10;
2. Download and use the dataset that is provided in the activity;
3. Closing and reopening WEKA and the Explorer perspective will reset every test option and algorithm parameters to their default value, so if you think that is the case, feel free to do it.

Please, let us know how it turns out.

Cheers,

Thales Vaz Maciel

Bacharel em Sistemas de Informação
Especialista em Sistemas Distribuídos
Senior Developer - HealthPlan Services, Tampa, FL, US
Desenvolvedor de Software - e-Core IT Solutions, Porto Alegre, RS, BR
http://lattes.cnpq.br/7166030596636868
F: +55 (53) 81030692

--
You received this message because you are subscribed to the Google Groups "WekaMOOC-general" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wekamooc-gener...@googlegroups.com.
To post to this group, send email to wekamooc...@googlegroups.com.
Visit this group at http://groups.google.com/group/wekamooc-general.
To view this discussion on the web, visit https://groups.google.com/d/msgid/wekamooc-general/a484270c-abe9-4e4d-8313-b7fd2c82bd27%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Thales Maciel]

Am I missing something?

In fact, you're missing the petalwidth attribute. lol

Please, follow my #2 instruction in the previous email and you'll be fine.

Best regards,

Thales Vaz Maciel

Bacharel em Sistemas de Informação
Especialista em Sistemas Distribuídos
Senior Developer - HealthPlan Services, Tampa, FL, US
Desenvolvedor de Software - e-Core IT Solutions, Porto Alegre, RS, BR
http://lattes.cnpq.br/7166030596636868
F: +55 (53) 81030692

[ngk]

Hi Thales,

Thank you very much for your reply. The significant change was actually the Weka version. (I had been using 3.6.9 and had been unable to upgrade to 3.6.10 without admin privileges until this morning - up to now the minor version difference hadn't impacted me but I should have heeded the warnings!). Incidentally iris.arff seems to be identical across all 3 of the following (datasets supplied with 3.6.9, datasets supplied with 3.6.10 and the download available with activity 3.1). Perhaps there is an implementation difference for the OneR algorithm between 3.6.9 and 3.6.10..

Great course by the way.

Thanks again,

Niall.