I performed a clustering using WEKA. I got 4 clusters and a confusion matrix. When I visualized the clusters, I thought I can understand the relationship between confusion matrix and the visual clusters. For example, if I see that there are 4 instances of "Good" in Cluster 3, I should see the 3 when I visualize them.
Can anyone tell me how to see the relationship between the confusion matrix given in the attached file (Class attributes.docx) and the visual clusters as shown in the "Question about clustering" image file.
you are plotting class vs cluster.
The colors in the graph are clusters colors.
Your confusion matrix says that you have 4 instances of the Fair class in cluster 3.
so you go to Fair on the X axis, and at the top (cluster 3 on the Y axis) there are four instances in light blue color, which indicates that they belong to cluster 3.
hope this helps,
you are plotting class vs cluster.
The colors in the graph are clusters colors.
Your confusion matrix says that you have 4 instances of the Fair class in cluster 3.
so you go to Fair on the X axis, and at the top (cluster 3 on the Y axis) there are four instances in light blue color, which indicates that they belong to cluster 3.
hope this helps,
From your confusion matrix : Attached another matrix in rows and columns
to interpret the attached table,
Columns are clusters and
Rows are classes.
To get cluster and the class from the look up table
f(clusterX, ClassY ) = No: of instances of cluster0 that belongs to class abnormal.
Eg: f(Cluster0, Class= 'Abnormal') = 3.
In the graph same no:of instances are visible in the area that belongs to the function arguments.
Hope it helps.
If a classification system has been trained to distinguish between cats, dogs and rabbits, a confusion matrix will summarize the results of testing the algorithm for further inspection. Assuming a sample of 27 animals — 8 cats, 6 dogs, and 13 rabbits, the resulting confusion matrix could look like the table below:
Predicted
class
Cat Dog Rabbit
Actual class
Cat 5 3 0
Dog 2 3 1
Rabbit 0 2 11
In this confusion matrix, of the 8 actual cats, the system predicted that three were dogs, and of the six dogs, it predicted that one was a rabbit and two were cats. We can see from the matrix that the system in question has trouble distinguishing between cats and dogs, but can make the distinction between rabbits and other types of animals pretty well. All correct guesses are located in the diagonal of the table, so it's easy to visually inspect the table for errors, as they will be represented by values outside the diagonal.
Assuming the confusion matrix above, its corresponding table of confusion, for the cat class, would be:
5 true positives 3 false negatives
(actual cats that were (cats that were incorrectly marked correctly classified as cats) as dogs )
2 false positives 17 true negatives
(dogs that were (all the remaining animals,
incorrectly labeled as cats) correctly classified as non-cats)
The final table of confusion would contain the average values for all classes combined.
Link: http://en.wikipedia.org/wiki/Confusion_matrix
Also Refer
Data mining with WEKA, Part 2: Classification and clustering
Link: http://www.ibm.com/developerworks/library/os-weka2/
PS..
You can save the model and it will be saved in arff format...From this file you can get details of which instances goes into which cluster....and about the number of clusters you can change it either through code or by changing the settings in the corresponding clusterer in the weka interface itself..... your result is well explained above...
The visualization is depicting the values given in confusion matrix indeed.
Ex: Bottom left corner has 3 dots ( Abnormal values classified as abnormal ). The confusion matrix also says the same. Similarly others. One small correction in that doc file that you uploaded.
All the correctly classified class count is along the diagonal. Hence the accuracy of the classification is 3+0+0+2= 5 / 86 . This is extremely poor. Hence clustering algorithm is a very bad idea to classify this data.
Hope this helped
Hi Tamilalagan,
Correction and Conclusion:
Your results are missing "Clustered Instances" section that has to be shown before your attached result. However, for you to have similar cluster assignments results from the plot that you attached, x-axis should refer to the "Instance_number", and y-axis should refer to the "Cluster".
On the other hand, conceptually, the standard "Confusion Matrix" is different from the matrix you showed.
HTH.
Samer
- Badges
- Science topic