Context: We run a collection of book clubs where members meet each week. Members are assigned to a specific club, but are free to join any club on any week. E.g., I may be assigned to the Tuesday club, but I can join the Monday or Thursday club if it better fits my schedule for a given week
Additionally, members are constantly leaving and joining the collective. E.g., member A might join for weeks 4-8, and member B might join for weeks 6-10. Thus, the specific composition of a club on any given week can vary significantly
We have an attendance table at the member-club-week level (e.g., on the week of 7/12, the Tuesday club had members A, C, D, F, and G in attendance)
I am trying to create a single quantitative measure of the consistency of co-members at the member level. If a member has the same co-members every single week that they attend, they should have a very high score. If a member has different co-members every time they attend, they should have a very low score.
My first thought was to use a market basket analysis to measure the co-occurence and confidence between each member. This gives us a quantitative score of how likely one member is to occur with any other member
However, where I am stuck is how to translate those two-member confidences into a single number at the member level which describes how often one members was in a club with consistent co-members
