Re: Understanding the confusion matrix and 'accuracy'

There is no model, it was only able to classify as true and deal with the
error.

If I have a dataset with 99T and 1F and I just classify everything as true
then I'm 99% accurate

On Sat, Nov 6, 2021, 2:59 AM Bob Matthews <[email protected]> wrote:

> Hi
>
> In an academic paper that I am trying to follow (with difficulty :)) they
> talk about a measure of accuracy
>
> being (TP + TN) / N
>
> TP = no. of correctly forecasted TRUE instances of the variable
>
> TN = no. of correctly forecasted FALSE instances of the variable
>
> N = total no. of instances in the testing instances
>
> But the confusion matrix I get using Auto-weka looks like:-
>
>     a   b   <-- classified as
>   376   0 |   a = TRUE
>   225   0 |   b = FALSE
>
> i.e. the predictions are 'TRUE' for every instance
>
> This I don't understand ?
>
> The paper gets accuracy above 80% but the above yields 62.5%
>
> and if the predictions are always 'TRUE' who needs a model ?
>
> What am I missing here ?
>
> Bob M
>
> _______________________________________________
> Wekalist mailing list -- [email protected]
> Send posts to [email protected]
> To unsubscribe send an email to [email protected]
> To subscribe, unsubscribe, etc., visit
> https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
> List etiquette:
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
>

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html