I have a dataset that consist of the following information: Person_1 (from node), Person_2 (to_node), Home_Room_Location (of person_1), Time_Spent_Together (treating this as weight).

I've been playing around with the dataset and found the following properties:

  1. If I exchange the position of the from_node to to_node, there is no disagreements on Time_Spent_Together (For example, if a -> b's weight is 5 then b -> a's weight is also 5 if b -> a exist in the dataset). However, only about 75% of the reversed data points can be found (For example, sometimes there is d -> e but not e -> d, but in the case where there is e -> d, the weight would match).

My question is: Since this is a relationship graph, where I assume relationship would go both ways (and the existing data agrees with me), is it a good idea to assume that this graph is undirected? Or should I stay on the safer side and assume it is directed since I don't know what could be in the unlisted datapoints?

  • $\begingroup$ A better question to ask is: What do you lose in terms of predictive power (whatever you're interested in) by assuming the graph is undirected? $\endgroup$ – Alex R. Jan 23 '17 at 19:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.