Tuesday, May 21, 2013

Update: Visualization


this week I added two new visualization methods to my simulation (see code on Github).

Updated visualization screen with the following options: Curve, Pie Chart and Bars

 The bar view

The pie chart

It is now also possible to view more than one sensor/constructs with each visualization option for comparison.

Here is an example of the curve viewing data for humidity and temperature:

I am currently updating my documentation with your hints and suggestions.
Thanks very much for your help!!

One hint from Thomas, which I don't really understand is:
... you should again think about your scenarios and the COMBINATION of constructs. ATM, you thought of combinations as mathematical equations. Combinations can also mean the visual combination of several graphs (constructs). This would also be a 'construct' that could be used itself ...

How can I visually combine sensors? If two constructs are nearly the same, storing e.g.  the mean value of them?
A concrete example (like a user story) would be great. I think I don't really understand what you mean with visual combinations.

Thanks a lot!

Cheers, Tanja

Monday, May 13, 2013

Preview Thesis


here you can find the preview of my thesis:

Thesis - Preview - PDF

Thesis - Preview - Microsoft Word

I would appreciate to know, if the direction I am heading to is right ... especially for the practical part (starting at chapter 7). Are some important parts missing?

Thanks a lot!
Cheers, Tanja

Sunday, May 12, 2013

Simulation GIT


after my simulation using the version control system SVN works, I implemented now the same with GIT.

I added a button called "Settings" on the simulation start screen.

When clicking the button, the settings can be chosen:

(I stored all the standard settings as default value).
When the settings for GIT are chosen, the text "GIT" is displayed on the top right corner of the start screen (instead of SVN) and when linking two constructs, a different windows service using GIT is started up.

I wanted to let the user chose, if the GIT version should store to the local repository or to the remote one on Github. I spend nearlly two full days browsing the internet and searching the source code: seems like it is not implemented to push to a remote repository while using the library Gitsharp.

So far, I experienced the following advantages/disadvantages with the two versioning systems:

I am currently working on the documentation of my practical implementation. I hope to be able to upload this document next week.

Cheers, Tanja

Sunday, May 5, 2013

Data Visualization

Dear readers,

this week I started working on the visualization part.

After the sensor called "Humidity" is checked and clicking the button "Show" ...

... my visualisation mask pops up:

  • Description
    Per default here stands the name of the selected sensor
  • Destination
    The destination, where the sensor file is stored
  • Data
    The current data in the sensor
  • Refresh
    The user has to click this button for refreshing the sensor data
  • Cancel
    Closes the mask
  • Show
    A new window showing the visualization opens

Coordinate system:

The coordinate system shows the timestamps on the x-axis (displayed interval = 1) and the actual data on the y-axis.

At first, I was not sure, how to visualize data with C#.
I started with online research. I found some self-made graphics and wanted to draw a graph on my own. But this started to get really complicated ... so I restarted browsing the Internet and came up with a open source library from the website "Code Project":

I implemented this code in my project. It works really fine now :-)
(see project called "Simulation" on Github https://github.com/tantschi/masterthesis.git).

At the beginning, I wanted to implement two different visualization methods.
I did some research on the Internet/books, but I don't know, which second method I should implement for my kind of data?
My sensor files contain just a timestamp and a value, but nothing else (e.g. 25.04.2013 09:17:59;93).
So I am not sure, if I should implement a second visualization or better try to focus on the processing/mining of the data (the storage in a versioning system).

Concerning my last post: I am now able to store the content of the linked sensors in a new construct-file.
I use another library from Code Project called "Math Parser".

It calculates the result out of my Link-text-field:

At the moment I link the two last lines in each text-file according to a given interval. I think it is quite hard to know which lines of two .txt-files with different sensor intervals belong to each other.