Journal on Policy & Complex Systems Vol. 2, Issue 2, Fall 2015 | Page 61

Journal on Policy and Complex Systems
mutation mechanism is introduced in the model . With mutation and a sufficiently long simulation time , it is possible to use a small quantity of agents to simulate agents with a sufficient number of possible combinations of parameters . With this in mind , in this model , agent parameters are initialized within a small range of all possible values , while a small portion of agents are initialized within a larger range . Mutation also exists in the hatch-and-die portion of the NetLogo simulation , ensuring that descendants have the opportunity to explore a larger space as well . Finally , Monte Carlo simulation is used to get the best simulation results with a minimal round of simulations .
4 . 7 Genetic Algorithm

The concept of “ Survival of the fittest ,” proposed by Darwin , can also be applied to the financial markets . According to the theory of evolution , descendants are born with behaviors that are likely better adapted to the changes in the environment they are being born into . With the benefits of the hatch-anddie mechanism in Netlogo , model designers can implement natural selection into the agent “ births ,” agent interactions , and agent behaviors . The best agents can have descendants that have similar trading rules , while the worst performers can be ruled out of the system ( in this case , when they have used up all their capital ). When agents get eliminated from the model , newly randomized agents with initial capital agents replace them to keep the number of agents constant , thus ensuring an active trading environment . The implemented genetic algorithm ensures that the total number of active agents in the simulation remains at the same level . The results of agents of different age are evaluated at the same time . Agents simply follow and learn from the richest agents within their radius .

4.8 Benchmark Agents

There are two additional agents generated in the model with the purpose

of helping the end user evaluate the model performance over time . These agents use the buy-and-hold strategy for BAC and S & P 500 , respectively . In other words , these agents buy a certain amount of BAC and S & P 500 stock shares in the beginning of the simulation . They hold onto those shares until the end of the simulation . Their cumulative returns are treated as the benchmarks for evaluating the performance of other agents in the same period .
58