Hi Yaseen,

Your classifier does not show good performance. Therefore, I recommend checking another metrics, like "Kappa statistic", before reconfiguring LibSVM's parameters, or even trying other classifiers. You may need also to check your data characteristics.

HTH.

Samer

Hi Yasseen

Recall is (true positives)/(true positives + false negatives). If this is 100%, you must have zero false negatives, so in that respect your classifier is working well.

Precision is (true positives)/(true positives + false positives). If this is 6%, you must have many more false positives than true positives, so in this respect your classifier is very weak.

In summary, if you don't care about false positives you have a strong classifier. If you do care then your classifier is very weak.

Tom

You may check the below attached paper which handles all the above classification parameters over a real life review data set.....I hope it will help you to clarify your doubt.....