Viz Lab Summer Grant 2006

2007 Reports 2006 Reports 2005 Reports 2004 Reports

Visualizations of Student Learning Style Data

James Allert Department of Computer Science

Project overview

I am attempting to develop a way of visualizing student learning by keeping track of student use of my online course materials and constructing graphics that illustrates patterns of interaction. Ultimately, I would like to correlate these patterns with conventionally-measured learning style data. The VDIL component was to consist of data-driven visualizations of the node usage on my network of web pages.

Use of the VDIL and its resources, Summer 2005

I put in numerous VDIL hours at odd times throughout the summer, especially late July and August attempting to develop the network simulation I needed. There were three main parts of this endeavor
+ implementing full X500 authentication on my website to enable the capture of data needed for this project
+ developing a visualization-relevant data file from the captured data
+ linking the data file to a suitable visualization program.

The authentication component did not get solved until August 12. I worked with Bruce Reeves, Scott Hollatz and Dan Burrows to figure out how to use the full X500 authentication system and guarantee security.

Sample data files were constructed in Microsoft Excel in the VDIL for testing with some of the visualization packages.

The bulk of my VDIL time was spent wrestling with TECPLOT 10 and SIGMAPLOT. These were the only two packages I could find that seemed to offer capabilities I might need. The learning curve is rather steep on TECPLOT and I borrowed the manual several times for the weekend. Ultimately however neither proved adequate for this project. However, the project goes on, with other resources I did discover and would like to recommend for the VDIL.

VDIL Findings and Recommendations

I mainly used the Curie machine, but from what I saw of the other PCs these comments apply equally to them all.

TECPLOT requires that a TECPLOT macro be installed into Microsoft Excel in order for Excel spreadsheets to be constructed that are compatible. With the graduate students in the lab we attempted several times to install the macro but were unsuccessful. This necessitated using straight text files, which wasn’t a horrible outcome, but added another step to my data processing. This TECPLOT macro should be installed in Microsoft Excel on the Curie machine so that Excel data can be read by TECPLOT. Overall, TECPLOT may have been able to simulate my network but the tradeoffs were a long leraning curve, considerable hacking around what it had available to produce what I wanted, and then visualization that could only occur through a proprietary TECPLOT viewer (requiring me to have TECPLOT installed on my machine in order to show it).

Reformatting my data files required some simple programming. However, there are no programming languages installed on Curie, except an old version of Visual Basic (Version 6.0). I recommend that Microsoft Visual Studio.NET compilers be installed on at least one machine in the VDIL so those with programming skill can use them. Instead, I shifted my files over to ub to use the compilers there, however, the ITSS ‘UMD Communications’ programs were not available on the programs menu, without QVTNet, or Tera Term it was difficult to get over to ub. A trial version of XWindows was available on the Curie machine but was limited to 2-hours and even at that, was unable to create a ub session. I recommend that the ITSS UMD Communications programs and a full, working version of XWindows be installed on the Curie machine.

As it turned out, TECPLOT and SIGMA PLOT proved inadequate for my tasks. Instead, I set about to discover several visualization tools that are more suited to what I am doing in hopes that the VDIL will consider aquiring them. The VDIL should invest in some good network visualization software. I think the VDIL would be more attractive to computer science and computer engineering faculty if it offered network simulation and visualization software. I recommend a free, open source product suite called Graphviz (www.graphviz.org).

Another product that I ended up using that absolutely should be in the VDIL is Microsoft Visio. Visio is Microsoft’s all-purpose visualization software. I found that it has templates for network visualization that can be directly linked to data in either Microsoft Excel spreadsheets or Microsoft Access databases. The shapes in the Visio drawings, in my case nodes on a network change properties in response to changes in the data. This allows for a dynamic model of network interaction—which is what I needed in the first place.

Summary

Overall, I made progress on about half of my project, but not on the visualization end of it. There were no VDIL programs that seemed to give me what I wanted, although the recommendations above should help that and make the VDIL more attractive to other researchers.

The VDIL seems to be an excellent facility for the production of multimedia programs, only adequate for scientific visualization, and inadequate for networking and process simulation and visualization. Given the importance of these in computer and engineering-related disciplines I think more attention should be paid to them.

So what progress have I made on the original visualization problem? Once fall semester got started I began monitoring the webhit data file I was amassing from my classes. It worked without a hitch. That meant that I could proceed to construct a program that reads and analyzes the file. This was done in Visual Basic over the course of the semester. The program depicts the web usage overall and when broken down by groupings such as gender, TA section, percentage of total points, and each of four different learning style categories. The results are animated to depict learners leaving a central learner space and visiting each of my web pages throughout the days of the semester. In addition to the visual web-access patterns that are displayed a comparative chart of hits/user is constructed as the program runs giving a summary evaluation of that variable in the form of a comparative line graph. Results are shown below (figures 1 and 2).

Figure 1: Screen capture of all user hits

Screen capture of hits by gender

In my VDIL presentation on Dec. 2nd I had a great time interacting with those in attendance. This is a strength of the VDIL and should be encouraged. Everyone seemed to have ideas about how to strengthen the visualization or where to go with it. I look forward to further association with the VDIL and hope to eventually develop this software to the point of it becoming a useful tool in the teaching of large classes.