Comments on: What is a good classification accuracy in data mining? http://www.dataminingblog.com/what-is-a-good-classification-accuracy-in-data-mining/ Data mining crossroads - research, applications, news, list of blogs and customized search engine about data mining. Tue, 07 Feb 2012 08:14:55 +0100 http://wordpress.org/?v=2.8.5 hourly 1 By: kapil http://www.dataminingblog.com/what-is-a-good-classification-accuracy-in-data-mining/comment-page-1/#comment-124547 kapil Tue, 17 Jan 2012 07:34:19 +0000 http://www.dataminingblog.com/?p=961#comment-124547 Great Article regarding Classification Accuracy. I want to add/suggest something useful formula in this... Classification Accuracy = Number of correctly classified testig examples/ Number Of testing Examples Great Article regarding Classification Accuracy. I want to add/suggest something useful formula in this…

Classification Accuracy = Number of correctly classified testig examples/ Number Of testing Examples

]]> By: Priya http://www.dataminingblog.com/what-is-a-good-classification-accuracy-in-data-mining/comment-page-1/#comment-105744 Priya Tue, 05 Apr 2011 02:46:22 +0000 http://www.dataminingblog.com/?p=961#comment-105744 Sir, I am doing my PG project related to credit scoring. I am obtaining the accuracy rate of around 70% and the precision rate of around 85% Is it the good classification range. Thankyou Sir,
I am doing my PG project related to credit scoring.

I am obtaining the accuracy rate of around 70%
and the precision rate of around 85%

Is it the good classification range.

Thankyou

]]> By: Sandro Saitta http://www.dataminingblog.com/what-is-a-good-classification-accuracy-in-data-mining/comment-page-1/#comment-77144 Sandro Saitta Tue, 27 Apr 2010 21:25:18 +0000 http://www.dataminingblog.com/?p=961#comment-77144 @Jeff: Thanks for your comment. Point 1) is particularly true in the online advertising domain where the click is a "rare" event. @Jeff: Thanks for your comment. Point 1) is particularly true in the online advertising domain where the click is a “rare” event.

]]> By: Jeff http://www.dataminingblog.com/what-is-a-good-classification-accuracy-in-data-mining/comment-page-1/#comment-76972 Jeff Mon, 19 Apr 2010 23:40:15 +0000 http://www.dataminingblog.com/?p=961#comment-76972 Hey Sandro, Interesting post. I think that this is one of the challenges I have faced when presenting a model (results) to a non technical audience. "You were only correct 70% of the time!?". You do need to explain the domain, the lift and the profit (if applicable) that comes from that improvement immediately when showing the accuracy. I would add a couple additional thoughts: 1) Just accuracy alone is not always enough - think of a *rare* event where prediction of 'false' is nearly always correct - but not useful. There, true positives, false positives etc. become key 2) In some domains, like database marketing, also with a rare event, you are less interested in classification accuracy per say as in the predicted probability and the lift of top deciles. My 2 cents Jeff Hey Sandro, Interesting post. I think that this is one of the challenges I have faced when presenting a model (results) to a non technical audience. “You were only correct 70% of the time!?”. You do need to explain the domain, the lift and the profit (if applicable) that comes from that improvement immediately when showing the accuracy.

I would add a couple additional thoughts:

1) Just accuracy alone is not always enough – think of a *rare* event where prediction of ‘false’ is nearly always correct – but not useful. There, true positives, false positives etc. become key

2) In some domains, like database marketing, also with a rare event, you are less interested in classification accuracy per say as in the predicted probability and the lift of top deciles.

My 2 cents
Jeff

]]>