Hello Everyone,
Hope everyone is good and doing well, I have developed a workflow for Decision Tree, to check the accuracy of Decsion tree model using Accuracy statistics (Scorer Node) and through ROC but here I am feeling difficulty how to plot ROC curve for this workflow, Please anyone can help, that will be great for me.
Thanking in Anticipation
aworker Thank you very much for your help in SVM workflow.
Hope you will help in this regard too
workflow is attached here
knime://LOCAL/MS%20Research%20October%202021/Decision%20Tree
Thanks for uploading the workflow and glad to help here too. I have two comments about it and a recommendation:
In the case of Classifiers based on Decision Trees and ensembles made of Decision Trees such as Random Forest, etc., you do not need to normalize the descriptors. This is because they treat every variable separately when deciding which one to use and how. This is not the case for neural networks (FFNN, SVM, RBF, etc.), where variables are all used together to compute distances and hence you need to normalize the variables to avoid bias towards the highly scaled ones. To summarize, it doesn’t harm to normalize but it is not needed.
The Decision Tree Predictor- node provides too with a Probability (Normalize Class Distribution) which you could use to generate your ROC curve:
Hi aworker, Sir I need a suggestion, I need to learn concepts of machine learning algorithms like (SVM, RF, ANN, Decision Tree, Logistic Regression, etc). what would you suggest from where I should learn concepts of these classifications as I am using these techniques un KNIME.
How to answer your question quite depends on your current level of knowledge on Machine Learning and on your background (Mathematics-Statistics / Informatics / Humanities / etc.). It would help if you could tell us a bit about your background. To start with a simple answer (assuming here a beginner’s level) , I would recommend to start with the KNIME documents available at the following link:
Many of them are provided free of charge and are very practical.
Thank you for your recommendations. Actually, I am doing MS in Transportation Engineering having Bachelors’s degree in Civil Engineering, and currently doing my Research in Machine learning classification techniques for accidents crash data. I am new to this field so your recommended books will help me. I hope you got some idea about my background. Actually, I am not developing some codes but checking the performance of different classification models using KNIME and WEKA software.
Please if you can give me some more recommendations or suggestions. I will start learning this book you mentioned
Thank you so much for your responses and guidance.
Respect
Thanks for your kind comments and for having validated the solution here too.
In this case, since you are familiar with mathematical notation, I would definitely recommend to follow from A to Z the extremely good tutorials by Jason Brownlee at his blog:
He is one of my preferred blog authors on Machine Learning. He is skilled, rigorous and very didactic. His books are really affordable in PDF format with plenty of examples and all of them with exercises are covered on his many blog articles as well. I would definitely recommend him to a MS Engineering student to start learning Machine Learning from scratch but with a good formal mathematical teaching.
His examples are not based on KNIME but on Python. Having said this, I believe it would be an excellent exercise for you to take these examples and try to translate them into KNIME workflows. Besides this, you could also take his Python code and directly integrate it into the KNIME Python nodes.
But as I said before, I would start first with the KNIME Press books to get really familiar with the KNIME environment and then the rest.
