Relationship weighting

[15onafuwadav]

Hello,

I am a 17-year-old student (Year 12 / 11th grade) writing an essay on link prediction algorithms and am new to neo4j so bear with me as I attempt to abysmally unfold my thought process in this question.

I am attempting to create a data set of a group of people with different relationships such as friends and co-workers etc. and then use the Adamic-Agar algorithm and preferential treatment algorithm available through the Graph Data Science plugin to test their prediction accuracy. I was wondering if it impossible to give the relationships weighting so that a friend is more important than a co-worker when the algorithm runs.

Also please inform me if there would be a better way for me to go about performing this experiment whether it would be using a different program or changing how I am thinking about the experiment.

Thanks for any help,
David

[tard_gabriel]

Hello @15onafuwadav

You can for sure add a weight property in your relations, but this will cost more time to execute your query.

[15onafuwadav]

Thank you @tard.gabriel for the reply

How much longer would execution take? And if not too much longer how would I go about doing it (syntax wise)?

[tard_gabriel]

Here with the free intro courses you can learn how to add a property to a node and a relationship and much more.

These are all the basics to introduce yourself to Neo4j, you can go further after that or even consult the neo4j cypher manuel.

[15onafuwadav]

Thank you very much!

[15onafuwadav]

Also @tard.gabriel is there a way that I can test the execution time of each algorithm I use giving a fairly precise numerical value (like syntax or software (I'm still relatively new to programming in general)).

[tard_gabriel]

You can learn much more than the execution time by preceding any of your Cypher query with the clause profile or explain.

[15onafuwadav]

@tard.gabriel
For my experimentation I will take an Instagram account with roughly 50 followers and use them to create my data set (weighting mutual friends as 1 and non-mutual friends as 0.5).

I will leave some of the links undone to test they accuracy of the algorithm to see if the algorithm will fill in the blanks and use this to create a confusion matrix. However on the Graph Data Science Plugin the algorithm is shown to only be able to predict the chance of one link at a time.

Therefore my question is will I have to predict the chance of each link individually or can I have the algorithm predict all of them and rank them in terms of their score? If so how will I go about doing this?