readme

来自「vc的源代码程序的基于最小特征向量的元素的集合方法的程序源代码」· 代码 · 共 54 行

Delete all the files except the following after the experiments. A copy of
all the scripts are kept into SCRIPTS subdirectory for safety.

Directory -> C4.5
Directory -> NBC
Directory -> SCRIPTS
Directory -> MISC
Shell Scripts -> crossVal        
Shell Scripts -> tryAll.CV
Shell Scripts -> discretize.CV
Shell Scripts -> sel
Shell Scripts -> classifier.CV    
Shell Scripts -> try.CV
Shell Scripts -> linkexe

To run the cross validation experiemnts you have to edit the "try.CV" script.
Please open the try.CV and change the parameters before conducting experiemnt.

To begin, issue the following command for "filename.data"

$ try.CV filename

Please note: the filename must be without extension

Also Note that the data along with the names file MUST BE in this CV 
subdirectory. While running the crossvalidation, a directory named 
"filename.CV" will be created within this CV subdirectory containing 
different crossvalidation results.

The summary of the cross validation result will be found in "filename.summary" 
file under "filename.CV" subdirectory. The "filename.summary" keeps the
following information

1st Column: C4.5 ErrorRate on continuous training data (bef. pruning)
            C4.5 ErrorRate on continuous test data     (bef. pruning)

2nd Column: C4.5 ErrorRate on continuous training data (aft. pruning)
            C4.5 ErrorRate on continuous test data     (aft. pruning)

3rd Column: C4.5 ErrorRate on discretized training data (bef. pruning)
            C4.5 ErrorRate on discretized test data     (bef. pruning)

4th Column: C4.5 ErrorRate on discretized training data (aft. pruning)
            C4.5 ErrorRate on discretized test data     (aft. pruning)

5th Column: NBC ErrorRate on discretized training data 
            NBC ErrorRate on discretized test data     

6th Column: Time taken to discretize

7th Column: Method used to discretize

In fact, for analysis the error rate for the training data should be 
considered