Researchers at the University of Auckland are leveraging artificial intelligence to assess the true value of companies based on profitability, efficiency, growth, risk, and more.
In a paper published in accounting research journalBusiness school academics Helen Lu and Paul Giersema have shown that machine learning algorithms can provide more accurate stock valuations than traditional methods.
Their machine learning method outperforms traditional models in valuation accuracy, and stocks identified as undervalued tend to rise in price, allowing investors to benefit without additional risk. you get the opportunity.
Not only that, but Lu and Geertsema have also developed machine learning algorithms to help experts identify competitors when traditional methods are not sufficient. This is especially helpful in a country like New Zealand where it’s hard to find obvious colleagues, Lou says. He has spent many years specializing in research applying artificial intelligence to solve financial problems.
This kind of machine learning technique is revolutionary because stock prices can be influenced by human biases because determining a company’s value involves many subjective choices, Lu said. increase.
According to Dr. Lu, industry professionals often evaluate companies against others in the same industry, but this process of determining which companies are peers is free from the effects of bias and evidence. This suggests that practitioners are strategically selecting peers to achieve the desired evaluation results.
“For example, in the case of a software company being evaluated, an industry professional would typically look for peers in the technology industry and ideally find a few that offer similar products, but that process is highly subjective. It’s a target,” Lu said.
“Because what exactly is the ‘tech industry’ and how do you define definitively whether some companies are on par? Maybe also consider pricing power and growth potential. Shouldn’t there be a need? Seasoned professionals often follow hunches that can be subject to personal prejudices.”
To minimize such problems, researchers trained and utilized so-called tree-based machine learning models. These models use company fundamentals to automatically figure out the best way to assign companies to different leaves in the tree. As a result, companies that are often assigned to the same leaf can be viewed as close peers with similar underlying principles. Conversely, companies that are rarely assigned to the same leaf have different fundamentals.
The researchers’ model analyzed a large sample of US common stocks listed on the NYSE, NASDAQ, and AMEX between January 1980 and December 2019. The final sample used in the machine learning model consisted of 1,811,785 company-month observations, corresponding to 16,201 companies. These models can be extended to stock markets around the world.
Not only did their approach produce more accurate valuations over time than traditional models across the enterprise, Lu said, the valuations also came closer to the company’s true value.
Dr. Helen Lu is a FinTech leader in the Master of Business Analytics program and Dr. Paul Geertsema teaches Financial Machine Learning and Data Analytics for MBA’s, both at the University of Auckland.
