Portfolio Selection and Trading Model based on Genetic Algorithms, K-Means Clustering, Fundamental Indicators and Technical Indicators

November 22, 2022

This thesis proposes a model capable of utilizing Evolutionary Algorithms and Fundamental Analysis for stock selection, this alongside a trading system based on Technical Analysis. Firstly a data scrapping program is created in order to obtain the financial statements of the S&P500 companies for every end of the month in the data range, allowing the calculation of various Fundamental Indicators. A K-means clustering algorithm is then used to classify these indicators among sector peers for every end of the month in the data range. A Genetic Algorithm will then search for the importance that should be given to each indicator using a previous three month stock price variation method as the fitness function. Three different strategies with different fundamental indicators are considered, one focused on growth, another on general performance and another focused on value companies. This stock selection algorithm is then tested under different portfolio compositions and alongside a trading system based on the EMA indicator, with the testing being done over the period of 2017-02-01 up to 2021-10-31. The results show that all strategies have potential, although the value and balanced strategies are the ones that most consistently beat the benchmark, yielding returns up to 119.4% in 2020, displaying their strength when recovering from market downtrends. The EMA trading system proved to be a useful tool as well, mainly when it comes to avoiding heavy losses and reducing volatility.

Authors in the community:

• 

  J

  João Quinas Guterres

  ist187043

Supervisors of this institution:

• 

  N

  Nuno Cavaco Gomes Horta

  ist13947

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

por - Portuguese

Instituto Superior Técnico