I started using Weka today. I am using a .arff file containing dummy information. Then, I use weka to get the results. However, the problem is the fact that I'm not sure what weka is doing to obtain the results. For example, using a Heoffding Tree, I got a correctly classified instances result of 75% and 25% on the incorrectly classified result; what exactly is 75%? What did weka do to get a 75% correctly classified result? Thanks!
0
votes
share step wise screenshots or code snippets you have. Anyway depending on the testing method(e.g., cross validation) weka reserves some percentage of your data for testing. e.g., suppose out of 100 rows in arff file, weka takes first 70 for training/learning. Remaining 30 for testing. When tested with 30 rows/instances of data weka got 27 correct and 3 wrong which implies 90 percentage accuracy.
– Sujith R Kumar
1 Answers
0
votes
Assuming you are using the Weka Explorer and its "Classify" pane, the evaluation result depends on your settings the in the "Test options" box.
The default setting here is "Cross-validation" with 10 folds. With this setting, your input data is split into 10 portions of around the same size. Weka then performs 10 runs of training and evaluating and gives you the averaged performance in the end. The process is explained in more detail in this stackoverflow post
