URGENT!! Need help with some work.

[Amy_Read1]

I am in desperate need of help with KNIME. I am using it for uni work and I’m getting really confused on what to do with my dataset.

I need to use kNN, SVM and Random Forest. However, I’m experiencing a few issues with data types.

For example, in my visualisations I found that two variables were valuable to use to explore further. I then created the streams but they just aren’t playing ball. I feel very overwhelmed with the amount of preprocessing options too.

Here is what my stream looks like:

We learnt about Naive-Bayes too but I don’t think it is useful for this dataset, which is why it is currently in there.

Dataset is below:
UCI Machine Learning Repository and click download.

If anyone would be so kind to give me advice or even better set up a call with me that would be amazing. I am so so stuck and no one is giving me help.

[sanket_2012]

Hi @Amy_Read1 ,
Welcome to the KNIME Community!

Could you please give us some information about the task? What is it about? What is the end goal? And what are the exact problems you are facing?

Thanks,
Sanket

[mlauber71]

@Amy_Read1 for this dataset I have set up some examples about how to do machine learning here. Maybe you can check this out.

Heart Disease - Machine Learning Case - Comparing Algorithms – mlauber71

For Algorithms that need normalization and just numeric data you might also take a look at the data preparation in the Metanode “BINARY_RPROP_MLP”. This will involve converting the string columns to numeric ones (you could also employ vtreat for that), to normalize the data and while doing that preserve the option to re-do this for new data. It does look a little bit complicated first but you can just use the example for other data, provided your target variable is named “Target”.

SVM with Parameter optimization:

SVM Exercise with Parameter Optimization – knime

You might want to discuss again what you would do with kNN. Is this to cluster the data. You can try and find an optimal number of clusters:

iteration over k values using variables in the K Nearest Neighbor node

Maybe like this. I have not checked if the results from the Iris sample data would be of any practical use in this scenario. [image] kn_example_loop_over_nearest_neighbor.knwf (68.9 KB)