※ 引述《LCL2 (唔~)》之銘言：
: It seems the there exists some mismatch in the log file format
: ex.
: 103/06/8(Sun)20:01:27,http://www.csie.ntu.edu.tw/~cjlin/libsvmzip,,103/06/8(Su
: n)21:31:58,http://www.csie.ntu.edu.tw/~cjlin/libsvmzip,61-223-238-147.HINET-IP
: .hinet.net,61.223.238.147,Mozilla/4.0 (compatible; MSIE 5.0; Windows 98; DigEx
: t),,
: the general attributes# is 7, however the above one repeat date and http
: address twice, which result in 9 attributes. This is the only one case
: found in whole log.
Well, finding this is already a good thing..
I don't know what happened, but maybe one day when I looked
at this file I wrongly deleted the remaining attributes of the
20:01:27 data
To make your life easier I decide to manually delete the 20:01:27 one.
: There exists some more:
: SV1),,
: ^^^^^^^^ the only words in one line.
I don't understand your question here.. Could you specify the line
number?
I saw some end with SV1),,
but ,, is ok. This means missing data
: What should we do with such "wierd" instances? Just eliminate them, or
: try to repair them? Anyway, if someone can give a brief explanation
: on the log format, I think it will be very helpful. Thanks a lot!
how to deal with such weird instances is part of your project.
ABout the log, you should have thought about how this software
was downloaded...
The cgi file generating the log is my htdocs/cgi-bin/libsvm.cgi