Portfolio management is the management of investments to meet long-term financial objectives.
Today, machine learning models and external data are used in order to help companies and individuals better manage, diversify, and maintain their assets and take on less risk for higher reward.
Adding machine learning (ML) to portfolio management is delivering concrete benefits for the manager. ML is incredibly fast and adaptable, lending itself particularly well to investment management.
ML takes the work of security analysts and strengthens it by:
Identifying particularly well performing equities within data sets
Making new forms of data analyzable
Reducing the negative effects of human biases on investment decisions
To create a good portfolio management model, you should have the historical data of all the securities you want to invest in as well as detailed financial information about public companies, including universal and verifiable financial information like quarterly to annual reports, 8-K filings, proxy statements, ownership filings, and many other forms.
Portfolio managers engaged in active investing pay close attention to market trends, shifts in the economy, changes to the political landscape, natural disasters, and news that affects companies as all this news affects investment sales and purchases.
ML can find patterns and meaning in the quarterly earnings calls of S&P 500 companies through the past twenty years. By comparing this information to stock performance, ML may generate insights applicable to statements by current CEOs.
Another example is examination of millions of satellite photographs in almost real time to predict Chinese agricultural crop yields while still in the field.
The main challenge that portfolio managers face is creating a portfolio with low risk and high return relative to the returns of other securities at the same risk level.
Another common problem is poor market liquidity combined with an uncertain investment range.
The Alephblog – Portfolio Management/
Deloitte: Artificial intelligence The next frontier for investment management firms
There are some techniques that produce significant improvements over traditional ones.
In estimating the likelihood of bond defaults, for example, analysts have usually applied sophisticated statistical models developed in the 1960s and 1980s respectively by Professors Edward Altman and James Ohlson (notably the Z and O scores). Researchers have found that ML techniques are approximately 10% more accurate than those prior models at predicting bond defaults.
Harvard Business Review: What Machine Learning Will Mean for Asset Managers