Generating Investment Strategies Using Multiobjective Genetic Programming And Internet Term Popularity Data

Abstract

Abstract Searching for stock picking strategies can be modelled as a multiobjective optimization problem. The objectives are mostly the profit and risk. Because of the conflicting nature of these objectives, we have to find Pareto optimal solutions. Multiobjective genetic programming (MOGP) can be used to find tree based solutions, using evolutionary operators. The advantage is that this algorithm can combine any number of inputs and generate complex models. Recent research shows, that the popularity of different terms on the internet can be used to enhance the models. This paper deals with a SPEA2 MOGP implementation, which uses Google trends and Wikipedia popularity to find stock investment strategies.