Controlling Risk Exposure in Periodic Environments: A Genetic Algorithm Approach

In this paper, we compare the performance of different agent’s investment strategies in an investment scenario with periodic returns and different types and levels of noise. We consider an investment model, where an agent decides the percentage of budget to risk at each time step. Afterwards, agent’s investment is evaluated in the market via a return on investment (RoI), which we assume is a stochastic process with unknown periodicities and different levels of noise. To control the risk exposure, we investigate approaches based on: technical analysis (Moving Least Squares, MLS), and evolutionary computation (Genetic Algorithms, GA). In our comparison, we also consider two reference strategies for zero-knowledge and complete-knowledge behaviors, respectively. In our approach, the performance of a strategy corresponds to the average budget that can be obtained with this strategy over a certain number of time steps. To this end, we perform some computer experiments, where for each strategy the budget obtained after a certain number of time steps is averaged over a large number of trials. To make the comparison fair, we first tune the parameters of each strategy. Afterwards, we compare the performance of these strategies for RoIs with different levels of noise and different periodicities.