Does meta-labeling add signal efficacy to trading strategies?

dc.contributor.advisorSilva, Raul Ikeda Gomes dapt_BR
dc.contributor.authorSilva, Thaíza Loiola
dc.contributor.coorientadorRibeiro, Ruy Monteiro
dc.coverage.cidadeSão Paulopt_BR
dc.coverage.paisBrasilpt_BR
dc.creatorSilva, Thaíza Loiola
dc.date.accessioned2023-06-03T18:13:24Z
dc.date.available2023-06-03T18:13:24Z
dc.date.issued2022
dc.description.otherMeta-labeling is a framework that consists of using a secondary machine learning model on top of a primary trading model with the goal of improving trading metrics such as Sharpe, maximum drawdown, etc. In this study, the primary model is a trading strategy based on MACD indicator, whereas the secondary embodies the Bollinger bands, volume and returns indicators and tests different meta labeling arrangements using Random Forest, XGBoost, Logistic Regression and Decision Tree. In all of them, the backtests metrics improved compared to the primary model alone. However, the machine learning models did not have an expressive performance which raises the necessity of testing the method furtherly in a context of a powerful trading system. Finally, this piece proposes suggestions on how to mitigate machine learning poor performance in the Meta-labeling trading context by using feature engineering based on computational power, exhaustive trading indicators combination, feature importance validation and hyperparameter tuning.pt_BR
dc.description.qualificationlevelGraduaçãopt_BR
dc.format.extent46 p.pt_BR
dc.format.mediumDigitalpt_BR
dc.identifier.urihttps://repositorio.insper.edu.br/handle/11224/5694
dc.language.isoInglêspt_BR
dc.rights.licenseTODOS OS DOCUMENTOS DESTA COLEÇÃO PODEM SER ACESSADOS, MANTENDO-SE OS DIREITOS DOS AUTORES PELA CITAÇÃO DA ORIGEMpt_BR
dc.subject.keywordsMeta-labelingpt_BR
dc.subject.keywordsmachine learningpt_BR
dc.subject.keywordstradingpt_BR
dc.subject.keywordstechnical indicatorpt_BR
dc.titleDoes meta-labeling add signal efficacy to trading strategies?pt_BR
dc.typebachelor thesis
dspace.entity.typePublication
local.subject.cnpqCiências Exatas e da Terrapt_BR
local.typeTrabalho de Conclusão de Cursopt_BR

Arquivos

Pacote original

Agora exibindo 1 - 2 de 2
Imagem de Miniatura
Nome:
Thaíza Loiola Silva_Trabalho.pdf
Tamanho:
2 MB
Formato:
Adobe Portable Document Format
Descrição:
Thaíza Loiola Silva_Trabalho
N/D
Nome:
Thaíza Loiola Silva_Termo Autorização.pdf
Tamanho:
457.64 KB
Formato:
Adobe Portable Document Format
Descrição:
Thaíza Loiola Silva_Termo Autorização

Licença do pacote

Agora exibindo 1 - 1 de 1
N/D
Nome:
license.txt
Tamanho:
282 B
Formato:
Item-specific license agreed upon to submission
Descrição: