Using RapidMiner to create networks

ashkan13ashkan13 Member Posts: 3 Contributor I
edited November 2018 in Help

Hi,

I have a huge database of soccer players for a time interval of 10 years. I am wondering if I can use RapidMiner to create a network of the players showing the players who are likely to be in a same team. I can say it is kind of a collaboration network. I really appreciate your kind comments.


Answers

  • haddockhaddock Member Posts: 849 Maven
    Hi there,

    Perhaps you could look at this exercise as 'frequent item' or 'association rules' mining, http://en.wikipedia.org/wiki/Association_rule_learning, there is an example ( 1_25 FPGrowth ) in the the samples provided to get you started.

    Good luck!


  • ashkan13ashkan13 Member Posts: 3 Contributor I

    Thanks for the reply. I've already tried association rules but the problem is that the matrix (word vector) which I get is very sparse and therefore it could not result in associations.

    I'm wondering if there is a tool in RapidMiner which can get the word vector and then just draw it as a graph.
  • Nils_WoehlerNils_Woehler Member Posts: 463 Maven
    If you just want to view the graph you can try to use the Transition Graph operator.

    Best,
    Nils
  • haddockhaddock Member Posts: 849 Maven
    Hi,

    I assume that when you refer to the 'word matrix' you a mean the association rules that are the results of the last operator in the example provided.  If you don't get enough rules, try setting the minimum support in the FPGowth operatora bit lower.
  • ashkan13ashkan13 Member Posts: 3 Contributor I

    By transition graph I think I need to give it the source node and the destination and then it draws the edge, am I right? If yes, it's not working in my case because I have a word vector of the relations between my tokens (players) which shows which players are teammates. Any ideas to draw this matrix?

    thx,
Sign In or Register to comment.