Practical Applications of Using Machine Learning to Forecast Market Direction with Efficient Frontier Coefficients

In Using Machine Learning to Forecast Market Direction with Efficient Frontier Coefficients, from the Summer 2023 issue of The Journal of Financial Data Science, Nolan Alexander and William Scherer of the University of Virginia introduce a strategy for outperforming the S&P 500 and other benchmarks. They forecast market returns by first applying a machine language model to three parameters describing the efficient frontier's position and shape. Next, they forecast conditional returns on individual assets using the CAPM and applying the properties of the inverse Mills ratio. Finally, they construct outperforming portfolios by combining the conditional forecasts of asset returns with a separately produced covariance matrix. "We took the three market coefficients and used them as input for a machine learning algorithm that allowed us to predict for one month whether sections of the market, like the S&P, were likely to go up or down. We then construct an optimal portfolio based on our predictions," said Alexander.