Ok i think i have it, thanks everyone for the help!
But there is an another problem.
How are you calculating the avg?
example:
----------- k-NN -----------
precision recall f1-score support
1.0 0.50 0.43 0.46 129
2.0 0.31 0.40 0.35 88
3.0 0.45 0.36 0.40 107
4.0 0.06 0.03 0.04 33
5.0 0.42 0.58 0.49 76
avg / total * 0.40* 0.40 0.40 433
so: (0.5+0.31+0.45+0.06+0.42) / 5 = 0.348* ~ 0.35* like i calculated it
in my avg part. Are you using some weights?
Class: 1
sensitivity:0.43
specificity: 0.81
ballanced accuracy: 0.62
precision 0.50
.
.
.
.
Class: 5
sensitivity:0.58
specificity: 0.83
ballanced accuracy: 0.70
precision 0.42
*avg total:*
sensitivity: 0.36
specificity: 0.85
avg ballance: 0.60
* avg precision: 0.35*
On 17 June 2015 at 16:06, Herbert Schulz <[email protected]> wrote:
> I actually computed it like this, maybe I did something in my TP,FP,FN,TN
> calculation wrong?
>
>
> c1,c2,c3,c4,c5=[1,0,0,0,0],[2,0,0,0,0],[3,0,0,0,0],[4,0,0,0,0],[5,0,0,0,0]
> alle=[c1,c2,c3,c4,c5]
>
>
> #as i mentioned 1 vs all, so the first element in the array is just the
> class
> #[1,0,0,0,0] == class 1, then in the order: TP,FP,FN,TN
> #maybe here is something wring:
>
> for i in alle:
> pred=predicted
>
> for k in range(len(predicted)):
>
> if float(i[0]) == y_test[k]:
> if float(i[0]) == pred[k]:
> i[1]+=1
> else:
> i[2]+=1
>
> elif pred[k] == float(i[0]):
> i[3]+=1
> elif pred[k] !=float(i[0]) and y_test[k] !=float(i[0]):
> i[4]+=1
>
> #specs looks like this: [1, 71, 51, 103, 208]
>
> sens=specs[1]/float(specs[1]+specs[3])
>
>
>
>
> if I'm calculatig
>
> sens=specs[1]/float(specs[1]+specs[2]) im getting also the recall like in
> the matrix.
>
> On 17 June 2015 at 15:42, Andreas Mueller <[email protected]> wrote:
>
>> Sensitivity is recall:
>> https://en.wikipedia.org/wiki/Sensitivity_and_specificity
>>
>> Recall is TP / (TP + FN) and precision is TP / (TP + FP).
>>
>> What did you compute?
>>
>>
>> On 06/17/2015 09:32 AM, Herbert Schulz wrote:
>>
>> Yeah i know, thats why I'm asking. i thought precision is not the same
>> like recall/sensitivity.
>>
>> recall == sensitivity!?
>>
>> But in this matrix, the precision is my calculated sensitivity, or is
>> the precision in this case the sensitivity?
>>
>> On 17 June 2015 at 15:29, Andreas Mueller <[email protected]> wrote:
>>
>>> Yeah that is the rounding of using %2f in the classification report.
>>>
>>>
>>> On 06/17/2015 09:20 AM, Joel Nothman wrote:
>>>
>>> To me, those numbers appear identical at 2 decimal places.
>>>
>>> On 17 June 2015 at 23:04, Herbert Schulz <[email protected]> wrote:
>>>
>>>> Hello everyone,
>>>>
>>>> i wrote a function to calculate the sensitivity,specificity, ballance
>>>> accuracy and accuracy from a confusion matrix.
>>>>
>>>>
>>>> Now i have a Problem, I'm getting different values when I'm comparing
>>>> my Values with those from the metrics.classification_report function.
>>>> The general problem ist, my predicted sensitivity is in the
>>>> classification report the precision value. I'm computing every sensitivity
>>>> with the one vs all approach. So e.g. Class 1 == true, class 2,3,4,5 are
>>>> the rest (not true).
>>>>
>>>> I did this only to get the specificity, and to compare if i computed
>>>> everything right.
>>>>
>>>>
>>>>
>>>> ----------- ensemble -----------
>>>>
>>>> precision recall f1-score support
>>>>
>>>> 1.0 * 0.56 * 0.68 0.61 129
>>>> 2.0 *0.28* 0.15 0.20 78
>>>> 3.0 * 0.45 * 0.47 0.46 116
>>>> 4.0 *0.29* 0.05 0.09 40
>>>> 5.0 * 0.44 * 0.66 0.53 70
>>>>
>>>> avg / total 0.43 0.47 0.43 433
>>>>
>>>>
>>>> Class: 1
>>>> sensitivity:*0.556962025316*
>>>> specificity: 0.850909090909
>>>> ballanced accuracy: 0.703935558113
>>>>
>>>> Class: 2
>>>> sensitivity:*0.279069767442*
>>>> specificity: 0.830769230769
>>>> ballanced accuracy: 0.554919499106
>>>>
>>>> Class: 3
>>>> sensitivity*:0.446280991736*
>>>> specificity: 0.801282051282
>>>> ballanced accuracy: 0.623781521509
>>>>
>>>> Class: 4
>>>> sensitivity:*0.285714285714*
>>>> specificity: 0.910798122066
>>>> ballanced accuracy: 0.59825620389
>>>>
>>>> Class: 5
>>>> sensitivity:*0.442307692308*
>>>> specificity: 0.927051671733
>>>> ballanced accuracy: 0.68467968202
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Scikit-learn-general mailing list
>>>> [email protected]
>>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>>
>>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>>
>>>
>>>
>>> _______________________________________________
>>> Scikit-learn-general mailing
>>> [email protected]https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Scikit-learn-general mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>>
>>
>>
>> _______________________________________________
>> Scikit-learn-general mailing
>> [email protected]https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>>
>>
>> ------------------------------------------------------------------------------
>>
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>
------------------------------------------------------------------------------
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
