Data Laboratory

[sumitd1234]

Well, I think I am asking too many questions here . Please do comment about my previous question as well. I was reading about the Information Visualization in social network analysis and I realized that parallel co-ordinates will be a nice to have info-vis feature for data- laboratory in multidimensional cases like friends on Facebook API can have numerical attributes like
i) Number of wall posts made on their profile page
ii) Number of groups which the user has joined
iii)Number of photos in which the user has been tagged
iv)Total number of friends
or non-numerical attributes like
Country, Gender, Time-zones etc.
There are two ways to create parallel-coordinates either 2-D or 3-D parallel co-ordinates. I am also planning to include dynamic attributes view with sparklines and scatter-plots. So does having 3-D parallel co-ordinates as a proposed feature will fit into the timeline or would it be good enough to have just 2-D parallel co-ordinates. Of course I will enjoy adding 3-D too later if not in SoC period. I am a bit lost in exactly defining the timeline. I would appreciate if you can guide me the feasible set of features to propose for the SoC period.

Thanks in advance.

[admin]

Hi,

Use JFreeChart in priority, as it is already included in Gephi for Statistics reporting.

But before asking on details, please start by thinking on use cases, major hurdles of the proposals, and ways to get rid of them.

[ragha]

Hi,

How about using clustering techniques for data visualization?
We could include clustering to create chunks of data based on its similarity. Kohonen maps could be used to visualize data in 2D maps.

I have experience with clustering techniques and would like to include this feature in Data Laboratory project.
Any thoughts in this regard?

Thanks,
Raghavendra

[mbastian]

Yes it could be interesting, but how long would it take on your planning? The program is short and the Data Laboratory roadmap is already well filled I think.

[ragha]

I could definitely implement K-Means clustering as i already worked on it before..
Should take like a week or so to finalize that

[mbastian]

Would be interesting to check this package as well:
Mondrian Interactive Statistical Data Visualization in Java

[sumitd1234]

Yes, I checked it out. It is nice .

[sumitd1234]

Hi,

I replied to the public review on GSoC interface. As I asked there I wanted to know that which type of UML diagram am I supposed to draw(class diagram or some other type). This is my first time with creating UML diagram .

[ragha]

@mbastian:
Missing value plots looks interesting. It could approximate the missing value (numeric) based on regression. Sounds like a cool feature to have. What do you think?

I haven't tested it yet...but am working on it.

-Ragha