Beadavi asked 4 years ago
Hello,
The network mapping script doesn’t always give the same results for cluster projection (single cluster) on the same batch of references with the same parameters, how to solve the problem please?
The names of the clusters are different and the number of references per cluster is also different.
Thank you in advance for your answer
Béatrice
Lionel Staff answered 4 years ago
Dear Béatrice,
Yes exactly, Louvain community detection will not classify exactly the nodes in the same clusters. It is particularly true for nodes which are at the periphery of the network or in between two clusters, or when the links are weak and values are low.
But it is supposed to be marginal in the network and affect only a really small proportion of nodes.
Are you using our new Louvain Resolution community detection algorithm?
I hope it helps
L
Beadavi answered 4 years ago
Hello,
Thank you for your answer. I have done some time ago a part of the analysis (Network mapping) according to a distribution of references by clusters and I would like to run a contingency matrix script with a cluster projection variable based on this old distribution. I have tried to re-run the nettwork mapping script many times with the same parameters to find an identical distribution but I can’t, the result is different each time.
How can I proceed? I have the gefx file and an excel corpus explorer extraction with the UT WOS keys and the clusters associated with the references, can they be used to restart this scipt?
Thank you in advance for your answer
Sincerely
Beatrice
aleabdoI’m marking this as solved since the follow-up question was moved to another topic. Cheers