Maximizing profit in uplift modeling through regret-optimal policy learning strategies