The primary goal of my master dissertation is to develop a guidance application for users who want to build a workflow on the KNIME platform. This guidance application should be intelligent and flexible to give users advice during the building process. I’ve checked the feature of Workflow Coach, but the guidance degree it can provide is not enough.
Then I want to have a look at the *.knwf file. This file is a compressed file that contains all the detail with respect to the corresponding workflow. Basically, it contains the following content.
My dissertation supervisor told me I can use data profiling method to analyse the dataset and find the patterns among this dataset. I find this guided exploration example is quite useful and I want to build this workflow via Python programme. My intentional scenario about this guidance application is after user uploaded their dataset, this Python programme will using data profiling methods to compute and analysis the feature of this dataset(e.g. find the most uniform histogram, most skewed histogram, most correlated columns), then generate a *.knwf file so users can simply import this file into KNIME to visualise the whole workflow. It would be better if the user could execute the new generated workflow to visualise the data and interact with those nice charts.
Thanks for your comment, it’s not easy work. I try to search for many resources that could give me a few hints but I failed. Your feedback let me know how big this project is and obviously, I cannot accomplish it within a month. But I’ll stick around in this plan for a while if someone’s comment could inspire me a lot.
in general KNIME nodes are XML files and the connections between them might also be interpreted as XML files and meta nodes.
You might aim for something like a technology demonstration where you create a simple workflow from a collection of nodes and create connections between them. Then you would discuss what are the challenges and present that as a:
discussion what would be necessary to create such a meta creator
create a small demo that would show how it would be done
That might be suitable for a masters degree. And also it might be interesting to see if you could create such a system with KNIME itself -)
As further inspiration I created a meta search using the XML files of KNIME:
The other example would be a backup strategy that would use an outside tool to handle KNIME workflows: