'Re: [Wekalist] Simple logistic output - interpretation'

[prev in list] [next in list] [prev in thread] [next in thread] 

List:       wekalist
Subject:    Re: [Wekalist] Simple logistic output - interpretation
From:       Mark Hall <mhall () pentaho ! com>
Date:       2013-01-31 8:25:11
Message-ID: CD308FC6.5A24%mhall () pentaho ! com
[Download RAW message or body]

[Attachment #2 (unknown)]

On 26/01/13 6:45 AM, "petra" <m885ra@yahoo.com> wrote:

>Hi,
>
>please can someone help me. I am new in this program (it is obligatory for
>me to use it), but I don't know how to interpretate some outputs.
>
>Here is how it looks like, pardon me, it is a long one, datas are written
>in
>my language (slovenian).
>I am confused because it made 2 classes, class 0 and class 1, and the only
>differences are or it is positive number or negative one. I do not even
>know
>what does the summary mean?
>And I am very confused about "Detailed Accuracy By Class" and "Confusion
>Matrix".

SimpleLogistic produces logistic regression functions for estimating the
probability of each class. In the two class case only one function is
actually necessary, since the probability for the second class is 1 -
probability of the first class. This is why the function for the second
class is the same as the first but with the sign of each coefficient
flipped.

The confusion matrix shows how many instances are misclassified by the
classifier and which classes they are mistaken for.

The detailed accuracy by class section shows information retrieval metrics
for each class. Take a look at:

http://en.wikipedia.org/wiki/Information_retrieval


Cheers,
Mark.

>
>
>Scheme:weka.classifiers.functions.SimpleLogistic -I 0 -M 500 -H 50 -W 0.0
>Relation:    
>napadi-weka.filters.unsupervised.attribute.Remove-R1-4-weka.filters.unsupe
>rvised.attribute.Remove-R29
>Instances:    112
>Attributes:   29
>              ovac
>              VELIKOST_L
>              OBDOBJE_IM
>              T_STANDARD
>              DOLÅ½INA_OB
>              VRSTA_OGRA
>              viÅ¡ine_ograj
>              min_viÅ¡ine_ogr
>              DOLÅ½INA_OG
>              ELEKTROOGR
>              Å T._Å½IC
>              VIÅ INA_OGR_el
>              MIN_VIÅ INA_el
>              ELEKTROMRE
>              VIÅ INA_MRE_elmr
>              MIN_VIÅ INA_elmr
>              FARMER_MRE
>              VIÅ INA_MRE_fmr
>              MIN_VIÅ INA_fmr
>              TOK_V_OGRA
>              IZOLATORJI
>              SPLOÅ NA_VZ
>              NOÃˆNA_OGRA
>              NAÃˆIN_VARO
>              VRZELI_V_O
>              OSEBNO_VAR
>              PES
>              Å T.PSOV
>              skode_diskr
>Test mode:evaluate on training data
>
>=== Classifier model (full training set) ===
>
>SimpleLogistic:
>
>Class 0 :
>-50.64 + 
>[ovac] * 0    +
>[VELIKOST_L=2] * 0.37 +
>[VELIKOST_L=3] * -0.62 +
>[VELIKOST_L=4] * -0.67 +
>[OBDOBJE_IM=1] * -0.26 +
>[OBDOBJE_IM=3] * 0.6  +
>[T_STANDARD] * 0.08 +
>[VRSTA_OGRA=druge] * -1.22 +
>[VRSTA_OGRA=elektroograja] * 0.43 +
>[VRSTA_OGRA=farmer_mreÅ¾a] * -0.54 +
>[VRSTA_OGRA=elektroograja+farmer_mreÅ¾a] * -0.34 +
>[VRSTA_OGRA=elektromreÅ¾a+farmer_mreÅ¾a] * 0.51 +
>[VRSTA_OGRA=elektroograja+elektromreÅ¾a] * -2.67 +
>[viÅ¡ine_ograj] * -0.01 +
>[min_viÅ¡ine_ogr] * -0.01 +
>[DOLÅ½INA_OG] * 0    +
>[ELEKTROOGR] * 0.1  +
>[Å T._Å½IC] * -0.13 +
>[VIÅ INA_OGR_el] * -0.01 +
>[MIN_VIÅ INA_el] * 0.02 +
>[ELEKTROMRE] * -0.22 +
>[VIÅ INA_MRE_elmr] * 0.55 +
>[FARMER_MRE] * 0.14 +
>[VIÅ INA_MRE_fmr] * -0.07 +
>[MIN_VIÅ INA_fmr] * -0.01 +
>[TOK_V_OGRA] * -0.12 +
>[NOÃˆNA_OGRA] * 0.14 +
>[NAÃˆIN_VARO=drugo_zapiranje] * -0.56 +
>[NAÃˆIN_VARO=ElektromreÅ¾a] * 0.83 +
>[NAÃˆIN_VARO=ElektromreÅ¾a_+_2_trakova] * -0.6 +
>[NAÃˆIN_VARO=Farmer_mreÅ¾a_5_Å¾ic+trak] * 1.96 +
>[VRZELI_V_O] * 0.13 +
>[PES] * 0.3  +
>[Å T.PSOV] * 0.22
>
>Class 1 :
>50.64 + 
>[ovac] * 0    +
>[VELIKOST_L=2] * -0.37 +
>[VELIKOST_L=3] * 0.62 +
>[VELIKOST_L=4] * 0.67 +
>[OBDOBJE_IM=1] * 0.26 +
>[OBDOBJE_IM=3] * -0.6 +
>[T_STANDARD] * -0.08 +
>[VRSTA_OGRA=druge] * 1.22 +
>[VRSTA_OGRA=elektroograja] * -0.43 +
>[VRSTA_OGRA=farmer_mreÅ¾a] * 0.54 +
>[VRSTA_OGRA=elektroograja+farmer_mreÅ¾a] * 0.34 +
>[VRSTA_OGRA=elektromreÅ¾a+farmer_mreÅ¾a] * -0.51 +
>[VRSTA_OGRA=elektroograja+elektromreÅ¾a] * 2.67 +
>[viÅ¡ine_ograj] * 0.01 +
>[min_viÅ¡ine_ogr] * 0.01 +
>[DOLÅ½INA_OG] * 0    +
>[ELEKTROOGR] * -0.1 +
>[Å T._Å½IC] * 0.13 +
>[VIÅ INA_OGR_el] * 0.01 +
>[MIN_VIÅ INA_el] * -0.02 +
>[ELEKTROMRE] * 0.22 +
>[VIÅ INA_MRE_elmr] * -0.55 +
>[FARMER_MRE] * -0.14 +
>[VIÅ INA_MRE_fmr] * 0.07 +
>[MIN_VIÅ INA_fmr] * 0.01 +
>[TOK_V_OGRA] * 0.12 +
>[NOÃˆNA_OGRA] * -0.14 +
>[NAÃˆIN_VARO=drugo_zapiranje] * 0.56 +
>[NAÃˆIN_VARO=ElektromreÅ¾a] * -0.83 +
>[NAÃˆIN_VARO=ElektromreÅ¾a_+_2_trakova] * 0.6  +
>[NAÃˆIN_VARO=Farmer_mreÅ¾a_5_Å¾ic+trak] * -1.96 +
>[VRZELI_V_O] * -0.13 +
>[PES] * -0.3 +
>[Å T.PSOV] * -0.22
>
>
>Time taken to build model: 0.47 seconds
>
>=== Evaluation on training set ===
>=== Summary ===
>
>Correctly Classified Instances          99               88.3929 %
>Incorrectly Classified Instances        13               11.6071 %
>Kappa statistic                          0.6533
>Mean absolute error                      0.1971
>Root mean squared error                  0.2949
>Relative absolute error                 52.2521 %
>Root relative squared error             68.1112 %
>Total Number of Instances              112
>
>=== Detailed Accuracy By Class ===
>
>               TP Rate   FP Rate   Precision   Recall  F-Measure   ROC
>Area 
>Class
>                 0.607     0.024      0.895     0.607     0.723
>0.942   
>1
>                 0.976     0.393      0.882     0.976     0.927
>0.942   
>0
>Weighted Avg.    0.884     0.301      0.885     0.884     0.876      0.942
>
>=== Confusion Matrix ===
>
>  a  b   <-- classified as
> 17 11 |  a = 1
>  2 82 |  b = 0
>
>Please can someone help me! Thank you very much!
>
>
>
>--
>View this message in context:
>http://weka.8497.n7.nabble.com/Simple-logistic-output-interpretation-tp267
>47.html
>Sent from the WEKA mailing list archive at Nabble.com.
>
>



_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/=
listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.=
html


[prev in list] [next in list] [prev in thread] [next in thread]
Configure | About | News | Add a list | Sponsored by KoreLogic