In my current project I need to build a model returning a vector of actions for each observation. I need a suggestion which statistical technique is used in general in such cases.
In a project, I have a dataset of about 100k observations, 7 vars (quantitative, qualitative). The variables describe a condition of a person. To each observation (each person) a sequence of 3 actions is applied so as to improve person's condition. Number of actions avaliable is limited (to 5), each action is of different average effectiveness and has different cost. A response variable is a percentage that reflects an improvement in person's condition after applying these actions in relation with the cost taken. (Of course, one action may somehow influence the effectiveness of the next action.)
The desired model should return a vector of 'optimal' actions for each observation. In my project, I have a learn sample (100k obs.) with a random sequence of actions applied to each observations and a response variable that reflects a real answer of 'how much' a sequence of random actions helped for this particular person. So question is: what statistical technique should be applied to find the patterns in data and get a model?
Please note I am an
[Update] Please note I need only a general suggestion, not a precise guideline.