Genetic algorithms applied to an evolutionary model of industrial dynamics

Abstract

• Implementation of a genetic algorithm in an evolutionary model of industrial dynamics using Netlogo. • The results corroborate with previous findings as the use of a machine learning method led to market dominance. • The results indicate that using a machine learning method do not necessarily improve the technological efficiency or the social welfare. • Actually, it was found a decrease in technological efficiency and social welfare when using a genetic algorithm, which contradicts previous works findings. In order to verify the effects of machine learning in a market structure, an evolutionary model containing firms that use a genetic algorithm to decide their investment in innovative R&D was developed. These firms share the market, with two other types of firms, those with a fixed rate of investment and those with random strategies. A model of industrial dynamics was implemented and simulated using several population distributions of the three types of firms. The availability of external credit and the length of learning periods were evaluated and their effects, in the market structure, analysed. The simulations results brought contrasting findings when compared to previous works, as it confirmed that machine learning led to market dominance, but the same did not occur when considering the improvement of technological efficiency and social welfare.