Hi,
I am new to KNIME and would like to build a DeepLeaning model on CIDDA-001 data set. ((Coburg Intrusion Detection Data Sets).
Any help will be appreciated.
Regards,
Mohammad
Happy to hear you started using KNIME and welcome in our forum!
Do you have some specific questions already?
If you first want to play a bit more with existing KNIME Deep Learning workflows I would suggest to visit our Hub here: https://hub.knime.com/search?q=deep%20learning
There you can download and check some examples.
The data set you are mentioning is only the data set correct? there was no existing machine learning model shared which could be used here?
Best,
Martyna
1 Like
Thank you Martyna,
I’ve tried to follow you tube videos and I came up with the workflow below;
However, I did not get good results.
Below is a idea of the data I have.
Please help if you can.
Hi @hammodeh_100,
Maybe this can be helpful a bit here:
https://towardsdatascience.com/deep-learning-tips-and-tricks-1ef708ec5f53
just out of my curiosity: is there a reason why you want to use DeepLearning for this data set?
Best,
Martyna
2 Likes
Thank you @ Martyna,
Which kind of data i need to do a deeplearning model. Sorry I am new here and learning. When i run the model, knime did not recognize all columns for learning.
Basically always try a more standard machine learning model before trying deep learning. The often work better and faster for basic tasks. My usual approach is to try Random Forest (RF) first because it doesn’t need any parameter tuning to detect if there is any signal in the data set. If with RF you can produce a somewhat usable model, then you can try a linear one first (eg. logistic regression) and if that is worse, try something more advanced (but which also needs more tuning) like xgboost.
Deep learning is for experts. DL models that are actually worth something usually have custom code in them which you can’t just create by clicking together “prebuilt components” even less so if you have what I feel 0 clue about what deep learning actually is and how it works.
Also I have a feeling this data set isn’t the easiest to work with as it sounds like some form of outlier detection. if you are doing this for learning purposes I would probably search for an easier data set.
3 Likes