Behavior: Say versus Do

[ldoolitt]

Hey All!

I've recently come across a problem as part of a personal project I've been working on. I'm looking for any guidance from a more experienced Data Mining / Pattern Recognition person about what algorithm may fit this problem type. I'm familiar with basics of classifiers, but I'm not sure how to approach this one.

So basically the problem is this:

Given...
...a finite activity set (1,...,N), a parent category of each activity, and a history of activities a person said they would do versus what they actually did, along with the time of day the activities took place, what the accompanying activities were, what the previous activities were along with the planned future activities...

...What is...
...the likelyhood that given a person says they will do activity A now that they will actually do activity A?

Example:
Person said they were going to do A and C at time t1
They did A and B at time t1
Person said they were going to do B and C at time t2
They did D and E at time t3

Person says they will do B and D at time t4
What is the likelyhood of that?

You can assume that each activity has a parent category also, like A and B belong to category K1, C, D and E belong to K2 or something like that.

Any suggestions would be great!

 

[Stella740pl]

Sounds like an interesting problem.
Is it, by any chance, a homework project?

 

[ldoolitt]

Hehe, oddly enough no it isn't. But now that I look at it I have formulated it kind of like a homework problem.

More of a thought problem than anything, but I was thinking it could be applied to medical Rxs. Like person A says he'll take medication K at time t and doesnt take it. Whats the probability that he'll take medication B at time t2? Could be important for alternative treatments; which one will have the best expected value, etc.