Fallout vs. False positive rate & false negative rate

classic Classic list List threaded Threaded
2 messages Options
rby
Reply | Threaded
Open this post in threaded view
|

Fallout vs. False positive rate & false negative rate

rby

Fallout is defined as the false positive rate but in Weka seems to the
FPR/2. Why this difference?

Also if you want to compare error rates, why in the visualization windows
the false negative rate is not provided as an option? Would be easy to add
it and very useful. Right now I do it looking at the percentages in the
confusion matrix but being able to see this as a graph would be really
great.

Thanks
Ric



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Fallout vs. False positive rate & false negative rate

Eibe Frank-3
To be honest, I have never looked at fallout before. In WEKA's code for TwoClassStats, it is defined as 1 - precision:

===========================

/**
 * Calculate the fallout. This is defined as
 * <p>
 *
 * <pre>
 * incorrectly classified negatives
 * --------------------------------
 *   total predicted as positive
 * </pre>
 *
 * @return the fallout
 */
public double getFallout() {
  if (0 == (m_TruePos + m_FalsePos)) {
    return Double.NaN;
  } else {
    return m_FalsePos / (m_TruePos + m_FalsePos);
  }
}

===========================

There appear to be some authors who define it the same way, e.g.:

Fu, B., Brennan, R., & O’Sullivan, D. (2009, December). Cross-lingual ontology mapping–an investigation of the impact of machine translation. In Asian Semantic Web Conference (pp. 1-15). Springer, Berlin, Heidelberg.

Regarding your other comment, FNR is obviously 1 - TPR, but you are right, it could be helpful to have it available as an option.

Note that you can save the data for a threshold curve from the window in which it is visualised.

Cheers,
Eibe

On Sun, Sep 1, 2019 at 2:23 PM rby <[hidden email]> wrote:

>
>
> Fallout is defined as the false positive rate but in Weka seems to the
> FPR/2. Why this difference?
>
> Also if you want to compare error rates, why in the visualization windows
> the false negative rate is not provided as an option? Would be easy to add
> it and very useful. Right now I do it looking at the percentages in the
> confusion matrix but being able to see this as a graph would be really
> great.
>
> Thanks
> Ric
>
>
>
> --
> Sent from: https://weka.8497.n7.nabble.com/
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html