User Tools

Site Tools


Term Project

Project teams

* Nathan LHOTE, Maxime SAVARO, Mohammed SENHAJI

* Haolin REN, Baubacar FANE (Tracking the trading bot)

* AUPETIT Jordan, LAUMOND Antoine.


* Jérôme Bouzillard et Marjorie Mersch.

* Jolly Solène, Lamoureux Adrien, Sommard Vivien

* Vincent Bocquel, Jérémy Turon

* Etienne Grandier-Vazeille, Benoit Barthès

* Simon Archipoff, Marie Gasparoux

* Nicolas Yvon, Morgane Badré, Maxime Eychenié

* Nicolas Spagnulo, Anne-laure Mesure

* Benjamin LETOURNEAU, Vincent WILMET

How will projects be evaluated?

It's worth studying this grid a bit, so you can get better grades – just imagine you have to grade your colleagues' work, or your own, after all.

Projects will be evaluated according to several criterion, each of them being assigned a grade [1, 5], interpreted in accordance with the criteria. The interpretation suggested here should not be taken too strictly, they are meant to give a flavor of what makes a difference between assigning a 2 or a 3, instead of a 3 or a 4, for instance.

Grades & possible interpretations
Criteria 1 2 3 4 5
Difficulty of problem addressed / questions asked Easy Requires some thinking but answer is straightforward Requires some thinking and average work Tricky, requires non obvious work Difficult, probably impossible to answer satisfactorily in given time
Implicit difficulty of dataset used Straightforward, not much work Required work but obvious solution Required work, inventive solution Needed non obvious combination of ideas and tools Difficult, probably needed too much time
Visualization (effort) Relatively easy, relied on ready-made solutions Good combinations of existing ideas and algorithms but could do better Some thinking, a mixture of ready-made and custom visualization, pretty good end result Astute and creative, pleasant and relevant Seems new, required non obvious work
Visualization (quality) Many things wrong, ended producing misleading graphical views Many things need to be fixed (wrong colormap, wrong mapping for size, wrong algorithms used, etc.) Overall OK, with only minor things to improve or add Very good impression, clean colormap choices, shapes, relevant layouts, etc. Additional aesthetics, almost close to professional graphics design
Visualization (creativity) None really Combining existing viz was a good idea Unusual use of existing algorithm and/or customization Astute solution, demonstrate mastery of ideas and techniques Clearly exhibit a creative mind
Visualization (relevancy) Didn't help much to solve the problem / answer questions Only gave a hint about a possible solution Was able to provide a non-surprising answer, with no real convincing arguments Clearly allowed to provide a firm answer to problem Brought additional insights on the problem
Written document (wiki) Poor, hardly readable or usable Contains everything it should, but needs to be reworked to be usable Globally OK but could be improved Good job, reads well Extremely well crafted, could even be published as a blog as is
Evaluation (discussion) Poor, confusing Everything seems to be there, but unorderly making it hard to grasp the work Globally OK, could improved (showing better images, emphasizing the right things, etc.) Good presentation, well balanced, lively, got all the information in the right order, straight to the point Entertaining, attractive, intellectually challenging discussion

What datasets?

I have presented the term project using data from the 7th Framework Programme (FP7 - funding research in Europe) listing all participations of organizations (research centers, private companies) to projects.

If you have a favorite dataset, do not hesitate to use it. Your dataset must not be too small, not too complex and somehow “clean” (this is not a data curation exercise). The web is after all plenty of “Open Data” …

Tracking the trading bot

  • I like the trading bot experience (a robot was given money and trades using automated algorithms, the deal is to predict when it goes bankrupt).
  • The data is readily available and modest (the bot has traded a bit less than 2000 times since it was launched).
  • Potential questions:
    • Does the bot perform trading based on any detectable patterns? Does the pattern evolve in time? How did it develop?
    • Is there any link between the share and/or companies the bot selects to bid on?
    • Can you provide evidence to support the prediction on the bot going bankrupt?
    • Etc.

FP7 – European money funding research

  • Here is a listing of all project participations. The data comes equipped with numerous attributes. Not all attributes are useful or usable.
    • Each entry is a project - organization pair.
    • The SO AGGREGATED attribute allows to assign each project to a strategic objective (a research area, let's say). It is thus possible to consider projects according to these strategic objectives. Note that a same organization may participate in multiple projects and thus be part of two or more strategic objectives.
  • Observe that the data itself allows to build a bipartite network where organizations link to projects but not directly to one another. It is however possible to link organizations directly (whenever they have co-participated in a project).
  • Potential questions:
    • What are the hub organizations? (The “big” actors in research), are these different according to different strategic objectives?
    • What are these organizations that cover multiple strategic objectives? Do they form some kind of interesting subnetwork?
    • Are there substantial budget difference between different strategic objectives?
    • Looking at collaborations between countries, do some European countries have special role for different parts of the world? What countries are getting more money from Europe? Does that relate with any specific strategic objective(s)?

There are tons of other datasets available

Just browse the web. The KONECT site is a good entry point.

/net/html/perso/melancon/Visual_Analytics_Course/data/pages/bordeaux20142015/termproject.txt · Last modified: 2015/01/06 16:03 by bpinaud