I would like to start by saying I've got nothing to do with statistics, but I need this for a machine learning task.
Now the issue: when I use Weka to try and predict a nominal value, the output contains "Correctly Classified Instances" and "Incorrectly Classified Instances" in percentages, which is a very easy way to understand just how efficient that particular algorithm is. why it has not happened on my results, how it will possible
When I use it to predict a numeric value, I get the following output:
Mean absolute error 0.3366 Root mean squared error 0.4036 Relative absolute error 74.0119 % Root relative squared error 84.6013 % Total Number of Instances 768
What does all this means? I've looked for all the terms online, but it goes beyond my comprehension. Can somebody help it for me please?
Thank you in advance!
Dear Talha,
The terms you are looking for are nothing but the performance metrics of the classifiers. To understand their meanings and equations, you may refer to a standard book, for instance, "Data Mining: Concepts and Techniques" by Han & Kamber.
When you are comparing values, it is more meaningful to compute the difference between true and predicted values (the error). The absolute error is the absolute difference between the true and predicted values. Mean absolute error is just the mean of these absolute differences over all instances. Mean squared error is the difference between true and predicted values to the power of two, and means squared error is just the mean of these values over all instances. Relative absolute error is the ratio between the error of the model and the error you would achieve if you predict the mean value of the target attribute for all instances. In a nutshell, you divide the (absolute or squared) difference of the predicted to the actual real value by the (absolute or squared) difference of the mean target value to the actual real value, and compute their mean.
Dear Talha,
The terms you are looking for are nothing but the performance metrics of the classifiers. To understand their meanings and equations, you may refer to a standard book, for instance, "Data Mining: Concepts and Techniques" by Han & Kamber.
Hi,
You' re performing a regression task (not a classification one); therefore, you are getting different evaluation metrics (BTW, you forgot highlighting the value of correlation coefficient produced by WEKA). Anyway, your results showed that your model is not an excellent one.
HTH.
Dr. Samer Sarsam
- Badges
- Science topic