I am using Cobweb clustering method on my dataset and evaluating it on trained data. I am receiving faulty output by the algorithm.
Part of the tree generated by Cobweb is:
node 0 [90]
| node 113 [6]
| | leaf 114 [1]
| node 113 [6]
| | leaf 115 [1]
| node 113 [6]
| | leaf 116 [1]
| node 113 [6]
| | node 117 [3]
| | | leaf 118 [1]
| | node 117 [3]
| | | leaf 119 [1]
| | node 117 [3]
| | | leaf 120 [1]
Whereas the evaluation output is:
113 1 ( 1%)
114 1 ( 1%)
115 1 ( 1%)
116 1 ( 1%)
118 1 ( 1%)
119 1 ( 1%)
120 1 ( 1%)
Technically, 113 should not be in the list since it is not a leaf node (like 117).
When I view the results after AddCluster filter, I see one instance assigned to cluster 113, which means that the tree generated is wrong.
There are other instances of the same problem in the output. Can somebody tell me what is lacking in my understanding? or if this is indeed something weird by Weka?
Part of the tree generated by Cobweb is:
node 0 [90]
| node 113 [6]
| | leaf 114 [1]
| node 113 [6]
| | leaf 115 [1]
| node 113 [6]
| | leaf 116 [1]
| node 113 [6]
| | node 117 [3]
| | | leaf 118 [1]
| | node 117 [3]
| | | leaf 119 [1]
| | node 117 [3]
| | | leaf 120 [1]
Whereas the evaluation output is:
113 1 ( 1%)
114 1 ( 1%)
115 1 ( 1%)
116 1 ( 1%)
118 1 ( 1%)
119 1 ( 1%)
120 1 ( 1%)
Technically, 113 should not be in the list since it is not a leaf node (like 117).
When I view the results after AddCluster filter, I see one instance assigned to cluster 113, which means that the tree generated is wrong.
There are other instances of the same problem in the output. Can somebody tell me what is lacking in my understanding? or if this is indeed something weird by Weka?