I've created a graph visualization for some bank transactions. It looks like the strength of the edges between nodes is based on the traditional IR (Information Retrieval) similarity/ranking models, i.e. cosine similarity and TF/IDF models, that are implemented in lucene. What I need is to strength the edges based on the amount or count of transactions (amount and count are two fields of each event captured by jdbc plugin of logstash from a database and stored in elasticsearch).
It seems that I should override or customize the similarity of elasticsearch for this index ... probably something like this. If this is the right path or there is another solution for it?
Thanks in advanced.