Markov Decision Processes formulation for stochastic and dynamic bank branches location problems

The optimization of investment policies in bank branches within dynamic and stochastic economic environment has become more and more important nowadays. However, it has not been generally formulated due to the randomness in markets and the complicated dynamics of economic growth. This paper formulates the stochastic and dynamic bank branches location problem as a Markov Decision Processes (MDP), and presents a policy iteration algorithm to obtain the optimal investment policies. Numerical examples demonstrate the effectiveness and efficiency of our formulation and algorithm. Furthermore, the formulation and algorithms have been embedded into an IBM asset called IFAO-SIMO, and they have been acting as the mathematical kernels of the asset optimization engine.