What methods can you recommend I use to assess the accuracy of the classifier I made?

Hi all, I am almost at the end of my master thesis. I developed a classifier based on pmoA gene. I was wondering what different methods I can do to assess the accuracy of the classifier I made. I am aware that there is a function in rescript for this purpose. However, it requires me to use a known database already. In my case, there is almost "not" known good classifier files for pmoA gene.

Any recommendations is highly welcome. Thank you very much.

Hi @pkmnsandy,

This is not true. Check out some of the tutorials listed on the github page. Some examples of interest are within the SILVA tutorial and this notebook.

Several features in RESCRIPt are specifically for the purpose of evaluating your novel reference database. You can read more about it here.


Hi Sandy,

Were you able to get your pmoA classifier using rescript? I am in the same position and would like to hear your comments on this.


Hello Stephanie,

Because it's been a few days (and a few years for Sandy), I thought I would chime in.

These days, I think the recommended way is to evaluate accuracy and investigate database structure with the newest version of RESCRIPt.

If you want to try something totally different, consider this bootstrap method from Robert Edgar. I'm not sure how automated this method is.

1 Like