how to measure classification errors using weka. can we take the value of RSME or etc to utilize for taken the classification rate?
You may follow the below links to measure classification errors...
http://www.damienfrancois.be/blog/files/modelperfcheatsheet.pdf
http://www.ijser.org/researchpaper/Comparative-Analysis-of-classification-Algorithms-using-WEKA-tool.pdf
http://research.ijcaonline.org/volume54/number13/pxc3882492.pdf
The following links provide some tutorials on how to measure classification errors in WEKA. I hope it helps.
http://machinelearningmastery.com/estimate-performance-machine-learning-algorithms-weka/
www.cs.waikato.ac.nz/~tcs/DataMining/Short/CH5_2up.pdf
Best!
Dear Muhammad,
Why you use weka, there are many softwares give you more reliability and accuracy than this software, such as
ENVI, PCI and other and if you need help I am ready to help you
my best regards
Hayder,
In your reply, you've claimed: " there are many softwares give you more reliability and accuracy than this software, such as ENVI, PCI". Therefore, could you give us an example that explains your claim (especially about reliability and accuracy)?
Dr. Samer Sarsam
I used WEKA for my experiments as authors of other papers had used it too and therefore it helps with comparisons etc. WEKA implements the leading data mining algorithms so is useful as a source for these algorithms, for example decision tree algorithms such as AdaBoost, MetaCost, C4.5 (implemented as J48), clustering algorithms such as SimpleKMeans, also leading algorithms from neural networks so its very useful. The output comes with many statistical measurements to test the accuracy of whatever algorithm is being run. However as WEKA is open source it is very easy to add to the output and implement any measurement you would like in addition to the ones already provided. The authors of the software produce a comprehensive text book which includes information with regard to the implementation of other algorithms/ new algorithms etc. I can recommend this text book as it will give users of any level good tips in using the software including adding code to the software. The authors are Ian H Witten, Eibe Frank and Mark A Hall and the book is called Data Mining Practical Machine Learning Tools and Techniques. I think that some data mining methods have different ways to measure classification errors that perhaps are specific to that algorithm but the book should explain major ways. Hope this helps.
Sanaa Fekry , your question is not clear; however, WEKA mainly uses the following error metrics:
- Mean absolute error
- Root mean squared error
- Relative absolute error
- Root relative squared error
Dr. Samer Sarsam
- Badges
- Science topic
- Similar topics
- Psychometrics
- Measurement