Federico Pepe's repositories
PortfolioStrategyBacktestUS
Master thesis project. The improved estimator of the covariance matrix of asset returns is employed to derive a new trading strategy based on a two-step procedure. First, it shrinks the asset universe via a subset selection, leaving only the most suitable assets. Then, it performs the mean-variance analysis. Back-testing is carried out in the U.S. stock market between 2018 and 2020. For comparison purposes, the code also implements also other strategies, such as the widely-used momentum strategy. The proposed technique is observed to deliver a very good and much more stable performance with respect to its competitors.
ImprovingMVPortfolio
This code tests the basic idea of my Master thesis. I propose an improved estimator of the covariance matrix of asset returns, employed in the computation of the minimum-variance portfolio. The main.py script tests the out-of-sample performance of this estimator, which is shown to deliver much better results than the sample covariance matrix and the equally-weighted portfolio.
PortfolioStrategy
Enhanced version of the Master thesis project. The original code has been enriched with a module that automatically downloads and stores new intraday data from Yahoo Finance, to serve as a real robo-advisor for investments. The code lets you choose between U.S. and European market, represented by stocks listed in the SP500 and STOXXE600 indices.
RNN-for-prediction-of-SP500
Prediction of SP500 up/down movements via Recurrent Neural Network