An Educational Management Problem with Continuous Signal Space

John E. Goulionis, V. K. Benos


Partially Observable Markov Decision Process  (POMDPs) have been suggested as a suitable
model to formalizing the planning of educational management.  In this paper, we discuss a specialization of POMDPs that is tailored to a frequently re-occurring type of educational problem, with five states (bad, moderate, good, very good, excellent), two teaching methods a traditional based to National program and a new education method based to the British system.
We extend the model of POMDPs with finite discrete signal space to a more natural model where the signal space is continuous instead of finite. We consider the significant and realistic problem with probability density functions for the signals to be uniformly distributed. We prove the piecewise affinity of the infinite horizon optimal utility function associated with this problem. To solve this problem we use a procedure that take advantage of  special problem structure, and we provide optimal policies to stochastic and dynamic decisions naturally arise in finding the optimal educational method.


OR in education; Markov processes; Bayesian statistics

Full Text:



eISSN 0975-5748; pISSN 0974-875X