Closed xurxodiz closed 12 years ago
Take good care of removing redundant variables (total number of deaths and deaths by enemy, for example) so not to mess with the results.
I need to correct myself: we don't need to integrate it with our architecture. We can generate the clusters (see #25) beforehand, then integrate the actual matching of user to cluster (#27) in the architecture for online classification.
Finished with commit c117c9310c7e32ef48c838a6f2dff878f2870e46. File ClusterGenerator.java
uses EM to read the arff file and output three clusters, saved in the Clusterer classfile clusters.dat
.
Implement the algorithm chosen in #23 in Java, integrating it with our current architecture.