Hi all,
I am working with weka (software for machine learning).
I have two sets of data, TRAIN and TEST.
1. The easiest question (I think). Is it possible to train modeland apply it to the TEST file but without removing an ID field?
It would be useful to join the predicted classification to the original data by using that ID field.
2. Would it be possible to train this model without removing the ID in the TRAIN file?
It would be useful to train and evaluate the model with the own training file.
I am trying it in the GUI of weka but my objective will be to do it via command line.
Thank you all in advance
You might. But the results wouldn't be according to your needs. IDs are just like conversation starters. They are there to inform you that okay this value exists. You need really need to remember them. Moreover, some of the IDs include gibberish numbers and alphabets that really don't contribute to the proper functioning of the dataset, and in machine learning, the main aim is to find the optimum values which contribute to the successful prediction of the solution. Also, you might also get errors while training the same. It's better to not involve gibberish stuff in the model, otherwise, the model is bound to fail.
Refer to your Q1, yes it is possible in Weka to use Train and Test with different datasets.
Thanks
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training