How Machine Learning is Transforming Portfolio Optimization (Technical Analysis)

The investment industry is undergoing a transformation that is largely attributed to technological advances. Investment professionals are integrating new technologies such as machine learning (ML) into the investment process, including portfolio construction. Many asset managers are starting to incorporate ML algorithms into portfolio optimization process, seeking more efficient portfolios than would be possible through traditional methods such as mean variance optimization (MVO). These trends call for a fresh look at how ML is changing the portfolio construction process.

Investors will benefit from a basic understanding of ML algorithms and the impact they have on their portfolios. Ultimately, the strategies used by asset managers to build client portfolios have a direct impact on the end investor. So investors should have enough awareness of these methods as they continue to gain in popularity. This article aims to provide an overview of role ML algorithms play in the portfolio optimization process.

Background

The term “machine learning” was first used by AL Samuel in 1959. Samuel conducted an experiment by training a computer to play checkers and concluded that the computer showed significant potential to learn. These results paved the way for further research on the subject and led to the development of increasingly powerful and sophisticated ML algorithms over the following decades. As a result, many industries, including investment management, have adopted these technologies in recent years.

ML algorithms are particularly useful when it comes to analyzing high-dimensional data or data sets with non-linear relationships, which is becoming more common with the rise of unstructured data and other alternative data sources. The two main categories for ML are supervised learning and unsupervised learning. With supervised learning, the ML algorithm detects patterns between a set of features (ie, input variables) and a known target variable (ie, output variable)(1). This is called a labeled data set because the target variable is defined. In unsupervised learning, however, the data set is unlabeled and the target variable is unknown. Thus, the algorithm seeks to identify patterns in the input data. Appendix 1 describes some of the common ML algorithms currently used by investment professionals.

Figure 1: Common machine learning algorithms in investment management.

Click to enlarge

Investment professionals expect new analytical methods to greatly disrupt the investment industry in the coming years. Respondents to a 2022 survey of more than 2,000 CFA Institute members predicted that new analytical methods such as ML will be the most important job disruptor in the next five to 10 years among respondents. Appendix 2 displays this result, along with other expected disruptors for job roles.

Figure 2. Factors predicted to significantly disrupt job roles in the next 5 – 10 years.

Portfolio optimization

The development of neural networks in the 1960s laid the foundation for many of the alternative portfolio optimization methods using ML. In addition, the emergence of “expert systems”(2) has led investment professionals to increasingly rely on machines to help solve complex problems. Some of the earliest uses of expert systems in finance include trading and financial planning expert systems.

The use of ML algorithms in the portfolio construction process has grown in popularity in recent years as investment professionals look for additional ways to improve portfolio returns and gain a competitive advantage. In particular, integrating ML algorithms into the portfolio construction process can address the challenges and limitations of traditional portfolio optimization methods such as MVO.

A major limitation of MVO is that it only considers the mean and variance of returns when optimizing a portfolio and does not account for skewness of returns. In reality, however, investment returns tend to be skewed. Specifically, research has shown that growth stocks have greater positive skew in their returns, on average, than value stocks. To account for the potential non-normality of investment returns, some investment professionals have chosen to construct portfolios using mean-variance-skew optimization models or even mean-variance-skewness-kurtosis optimization models. These models, however, result in multi-objective optimization problems. ANNs can effectively create mean-variance and skewness-optimal portfolios to address this limitation.

Another shortcoming of MVO is that it prevents investors from expressing their views on the future performance of assets. An investor, for example, might expect bonds to outperform stocks over the next six months. The Black-Litterman (1992) model allows investors to incorporate these insights into the portfolio optimization process. An alternative approach is to integrate the Black-Litterman (1992) model with ANNs, which has the potential to generate high returns relative to the benchmark without taking on excess risk.

The inputs to MVO are sensitive to measurement error, which is especially true for estimates of expected return. Thus, MVO has the potential to produce “optimal” portfolios with poor performance. Inverse optimization can be a useful alternative to develop more accurate estimates of expected return. Investment professionals can use these improved estimates as inputs into traditional MVOs to generate more efficient asset allocations. Investment professionals can also use ML algorithms to predict stock returns and incorporate these estimates into MVO. Alternatively, a recent study developed an improved portfolio optimization approach, which consists of using a correlation shrinkage parameter to improve the estimated Sharpe ratios and then creating optimal portfolios based on these estimates.

Finally, a major challenge in portfolio optimization is the estimation of the covariance matrix, especially for high-dimensional data. LASSO models can address this challenge by producing more accurate estimates of the covariance matrix than traditional methods, which is a critical input to MVO.

ConCluSIonS

What are the implications of these trends for investment professionals? Clearly, the investment industry is rapidly evolving in response to new technologies. Investment professionals anticipate that new analytical methods such as ML will significantly disrupt job functions in the coming years. As a result, practitioners are beginning to integrate ML algorithms into all areas of the investment process.

Many asset managers seek to gain a competitive advantage by creating portfolios with higher returns for a given level of risk (ie, higher Sharpe ratios) by integrating ML algorithms into the portfolio optimization process. In addition, ML algorithms can overcome many of the challenges and limitations of traditional portfolio optimization methods, which has led investment professionals to seek more efficient methods of portfolio construction. Investors will benefit from greater awareness of these trends to better understand the impact of new optimization methods on their portfolios.

(1) In some cases, the dataset may have more than one target variable.

(2) An expert system describes a computer program that can solve a complex problem normally solved by human experts. See: Expert System | AI, Knowledge Representation and Reasoning | Britannica

Disclaimer: Please note that the content of this site should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute.

Original post

Editor’s note: The abstracts for this article were chosen by the editors of Seeking Alpha.