How to use MOSES for supervised Classification?

[Vishnu Priya]

Hello All,

I read http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0085733. MOSES is being used for predicting the risk of suicide.  The classification of different groups were done based on the following steps. 

Bag of words -> feature selection-> Model building -> k fold cross validation -> ensemble of representations

I also have a similar classification problem. I want to classify economic activities based on ISIC code (International Standard Industrial Classification). 

Example:

 training Set:

Words                                                                                 ISIC Code

{mining, coal, conveyors, steam engine, thermal }  -          1111

{food, fiber, silk, drugs, flowers}                               -          2222

{hydraulic jack, brown coal, retreat mining}             -          1111

{wool, orange, vegetable, cocoa}                             -          2222

So after training, when i give some economic activity, MOSES should be able to classify it. 

But i don't know how to start. i could find only simple tutorial examples like AND/OR gate.  Could any one guide me ?

Thanks and Cheers,

Vishnu

[Nil Geisweiller]

Hi,

On 05/11/2017 06:48 PM, Vishnu Priya wrote:
> But i don't know how to start. i could find only simple tutorial
> examples like AND/OR gate. Could any one guide me ?

I don't understand in details your example. Maybe suffice to say that
you can use MOSES to learn

1. boolean models, B^n -> B
2. numerical models, R^n -> R
3. numerico-boolean models, (B union R)^n -> (B union R)
4. simplistic programs to control an agent

It excels at learning boolean models, not so much for the rest. You can
find examples in
https://github.com/opencog/moses/blob/master/tests/moses/MOSESUTest.cxxtest

Nil

>
>
> Thanks and Cheers,
> Vishnu
>
>
> --
> You received this message because you are subscribed to the Google
> Groups "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to opencog+u...@googlegroups.com
> <mailto:opencog+u...@googlegroups.com>.
> To post to this group, send email to ope...@googlegroups.com
> <mailto:ope...@googlegroups.com>.
> Visit this group at https://groups.google.com/group/opencog.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/opencog/91dba829-1cbb-4d2a-914c-26d4edaaba8f%40googlegroups.com
> <https://groups.google.com/d/msgid/opencog/91dba829-1cbb-4d2a-914c-26d4edaaba8f%40googlegroups.com?utm_medium=email&utm_source=footer>.
> For more options, visit https://groups.google.com/d/optout.

[Vishnu Priya]

Thanks Nil for the reply. 

I don't understand in details your example.

  

Let me try to define my problem again. It is a classification Problem. I have group of words that belong to certain categories. 

Eg: group of words that belong to category "bakery Products". Another group of words that belong to "vegetable"

Let me assign,  bakery Products- 1

                         Vegetable- 2

Words                                                     code 

{cookies, crackers, waffels, rolls}              1

{tortillas, pies, tarts}                                   1

{potato, carrot, lemon}                               2

{turnip, cabbage, mangoes}                      2

.

.

. etc

I want to train the above data with MOSES and then when i give certain words, it should predict the correct code for it. 

    Sweet potato  -?

      cakes            -?

Then MOSES should predict 

    Sweet potato  -2

      cakes            -1

i think, it is a linguistic driven prediction model. 

I read a paper, in that also words were taken and  classified according to certain categories USING MOSES and it was mentioned that accuracy was 65%.  eg.. Like below:

Words                                               Category

{agitation, extreme, unstable}  -        suicidal group

{unique, practicing, Neut}        -         Psychiatric group

{Plasma, neutral, vessel}         -         Control group.

So my problem also could be solved using MOSES? 

Thanks, 

Vishnu