Many of the popular techniques in Artificial Intelligence (AI) have been used in real-life applications since the 1980s, including machine-learning (ML) techniques.
In fact, as a high school student, I first came across ML techniques while studying calculus and linear algebra. During my undergraduate and graduate years, I used ML as part of my tool kit. Here at Intech®, as a quantitative manager, it is only natural to employ AI as a research tool, alongside more traditional quantitative methodologies.
AI is generally implemented as a model that ‘learns’ from training data and makes predictions based on new data. The accuracy of AI is based on the suitability of the algorithm and the quality and wealth of the data that is used for training and for making predictions.
At Intech®, our philosophical approach is to focus on the stable features of the markets and to develop a scientific understanding for their behavior. For this reason, we do not use AI directly to make investment decisions, especially in an unsupervised manner.
Yet, we consider machine-learning techniques as complementary to more conventional approaches: machine-learning techniques are helpful in generating research ideas, while conventional techniques are vital for validating these ideas and discovering the underlying explanation.
Machine learning should be considered a category of AI. In general, AI may be viewed as machine thinking on a range of tasks; whereas, machine learning is thinking about a specific task and making predictions. For example, a general-purpose robot thinking on its own is AI; a self-driving car is machine learning.
Over the past 15 years at Intech®, we have applied a variety of machine-learning techniques to assist in our ongoing research on market structure (including volatility and correlations as a source of returns, risk premia, and trading costs), as well as for monitoring our portfolios and developing improvements to the investment process.
We have used machine-learning techniques such as neural networks, fuzzy logic, support vector machines, dimensionality reduction, clustering, genetic algorithms, and ensemble learning for generating research ideas.
Our application of machine-learning techniques has been facilitated by extensive and high-quality proprietary datasets, such as a carefully curated global stock-returns database and a comprehensive record of Intech’s trades. We have also benefited from long-term academic collaborations.
In each application of ML, the findings may represent over-fitting or anomalies in the data. Even if they are genuine, they may identify transient patterns that are not likely to persist.
As such, we plan to continue relying on the scientific method while employing AI techniques as a research tool, and to continue investing resources in exploring and adopting developments in the field. To learn more about the role of AI in the field of finance please read our most recent paper: Artificial Intelligence in Finance.
How to Use AI:
- use as general a model as possible (so that the ever-present biases are reduced)
- use as broad a training set as possible (so that more significant cases are captured)
- have an independent reasonableness check (so that spurious patterns are weeded out)
This information is intended to be educational and is not tailored to the investment needs of any specific investor, nor is it an endorsement or recommendation for any particular security or trading strategy. You should not rely on this information as the primary basis for your investment, financial, or tax planning decisions.
