Risk induced k-min search algorithms: An experimental perspective

In this paper we address the k-min search problem under the risk-reward framework. In a k-min search problem a player wishes to purchase k units of an item, with the objective to minimize the total buying cost. In Computer Science this problem is studied under the competitive analysis paradigm. Lorenz et al. and Iqbal and Ahmad proposed algorithms (namely LPS and Hybrid respectively) to solve the k-min search problem under the competitive analysis approach. However, the main drawback of the competitive analysis is the assumption that the input is always a worst-case and thus resulting in risk-mitigating algorithms. We consider a scenario where an investor will like to introduce risk in his decision making and test algorithms on real world data by introducing risk in the decision making criterion. We observe that Hybrid performs better than LPS.