We are exploring how to set up the KNIME Server Web Portal for use within the classroom.
We have business school students who, at best, would be data analysts (not data scientists). We would like the students to access pre-built workflows, allow them to tweak specific input variables, and then ask them to make business recommendations based upon the results.
We do NOT want students to install KNIME or even see the full detail of the workflows. We want students to access simplified and limited parts of the workflow through their web browser.
But we DO want each student to be able to run their own instance of each workflow. Each student should be able to set their own input variables without interfering with anybody else. Hidden randomized parameters might be used to vary each workflow instance so that the output from each student’s workflow would be slightly different.
We would also like to avoid heavy involvement from the IT department. Ideally a professor would be able to develop a workflow (with the web portal hooks) then post it someplace for the students to access in the same, simple, way a professor might post an Excel spreadsheet.
Thanks for posting, this sounds like a really interesting use-case for KNIME Server and WebPortal!
For clarification: Should students also be able to change parts of a workflow, e.g. reconnecting nodes, inserting new nodes, deleting nodes?
Every invocation of a workflow will be a dedicated job with the parameters set via the WebPortal. Individual jobs will not interfere with each other, so this will not be an issue. The randomization per job is a slightly different story depending on what exactly should be randomized. Do you already have an idea concerning this? To me this mostly sounds like data shuffling or some such.
Let’s clarify above questions and get work on this started!
We’ve developed a “Market Simulation” agent-based model that runs on KNIME as a collection of nodes (see www.scientificstrategy.com). We use these nodes to build workflows that simulate various microeconomic phenomenon and business school case studies. The business questions we would ask students might include:
Is this an attractive market to enter?
If you enter, how much capacity should you produce?
What price should you charge?
How should you react if a competitor entered with a more aggressive price?
What features and product-line extensions would help you grow market share?
In many cases, the students would follow along the story from an actual historic case study, making decisions along the way in a “choose your own adventure” fashion.
So to answer your questions:
Should students also be able to change parts of a workflow, e.g. reconnecting nodes, inserting new nodes, deleting nodes?
In the beginning, we’d like to give students a simple menu-like user interface where they could select “enter_market=yes / no”, “product=A / B / C”, “price=$”, “feature=x / y / z”. Later, it would be nice if students were able to add their own nodes to a workflow that would help them make certain measurements, like the “price elasticity ratio”, but this would be optional and I can see many students preferring to copy-and-paste a table into Excel for more detailed analysis.
Do you already have an idea concerning the randomization per job?
We’d like to alter the sensitivity of each student’s market so student’s can’t simply rely on their classmates for the analysis. For example, Student_001 might have a very sensitive market in which sales are highly dependent upon price, while Student_002 might have a less sensitive market in which product quality is a more important factor. But this could also be a menu-option, and we could assign students to a particular market condition, telling them in advance “you are in market=G”.
This is already possible using Wrapped Metanodes. They can be published as steps of a wizard in the WebPortal version of a workflow. You/the lecturer would be developing those kind of workflows and, subsequently, deploy them to a KNIME Server that the students can access.
This would currently involve some manual labor. You could use a drop down to let the students select an additional node from a limited set of nodes. The limitation is required since you have to decide based on the drop down which branch to activate in the workflow. But that’s really just a workaround and we are working on improving the situation.
If you make this the first step in a workflow and students have to select their market themselves, you should be able to cover your bases with only one workflow of which each student can start their individual jobs. If you/the lecturer wants to select e.g. market sensitivity for the students, we might have to get a little more creative.
