Machine learning, a revolution for investments

Nowadays, daily life is influenced by actions and recommendations based on machine learning, such as movie suggestions on digital platforms, facial recognition, chatbot assistance, among others. This same technology is increasingly being used in investments to improve decision making due to the added value it offers compared to traditional programming.

Automatic learning or machine learning (ML) is not a recent topic; one of the most important precedents was during the 1950s, when Alan Turing raised the question of whether it was possible for machines to think. At that time, this technology was not booming due to the lack of information and computational power compared to what it has today.

Machine learning can be defined as a branch of artificial intelligence where the computer is capable of learning by itself, analyzing and recognizing patterns among a large amount of data. In investments, this technology is useful for understanding economic-financial problems and improving the analysis of this type of variable.

To understand what the added value of this technology is and why it is revolutionary, it is worth comparing it with traditional programming. Both involve data, decision rules and results (responses) applied to the information. The difference is that in classical programming the human defines the rules and analyzes the results obtained, while in machine learning the human enters the information together with the desired answer and it is the ML algorithm that deduces the pattern (rules) to obtain that result by learning from the data provided. An example applied to investments where the objective is to obtain profits in a portfolio would be the following:

Classical Programming: An investor has the belief that stocks show positive returns when economic variables such as consumption, investment and employment have favorable growth. The investor can do a retrospective analysis to see what his profit on the portfolio (result) would have been if he had invested in stocks when the economic variables showed a positive trend (decision rule). This is how the human is defining the rules to the data and analyzes the results obtained.

Machine learning: Using this perspective, the investor provides the ML algorithm with the economic variables (data) and the days or months in which the stock had positive returns (desired response). Subsequently, the algorithm learns from this information and provides the human with the decision pattern that is most likely to acquire profits on their portfolio. In this case, a possible pattern identified by the algorithm is that stocks are more sensitive to investment and consumption compared to employment, or that the market anticipates the behavior of economic variables, among other possible sets of rules. In short, the human provides the data and the desired result so that machine learning deduces the decision rules that may vary depending on the type of algorithm, training and validation used.

One of the biggest benefits of machine learning is the flexibility and customization it provides. Just as advertising on social networks is in accordance with the preferences of each individual, in investments it is also possible to obtain personalized models and solutions for the different economic-financial variables and types of clients.

Although machine learning is not a complete substitute for traditional programming, it is a fact that it is revolutionizing the financial industry due to the differentiation it offers to complement investment analysis and thesis. This is why more than 75% of financial services companies are investing resources in developing tools through machine learning. Investing is an opportunity to grow wealth and you can take advantage of the wide range of investment options that exist, such as funds, as they are increasingly accessible to any type of investor.

*Julio Sánchez is Executive Director of Quantitative Models at BBVA Asset Management.

julioalberto.sanchez.trujillo@bbva.com