Researchers Explore Use of Machine Learning in Predicting Fund Performance

A couple of research papers evaluate the uses and value of artificial intelligence and machine learning in predicting mutual fund performance. In one paper, researchers from the London Business School and Barcelona School of Economics, among others, investigate whether investors can use machine learning combined with publicly available data to construct portfolios of mutual funds that deliver positive returns net of all costs. The researchers posit that machine-learning methods can use information contained in multiple fund characteristics to select funds that earn economically and statistically significant positive risk-adjusted returns net of fees and transaction costs. They write that, among other advantages, their methods have the ability to identify and exploit nonlinearities and interactions between multiple predictors. In contrast, “linear forecasting models can help investors only to avoid negative alphas. Therefore, our results demonstrate that, employing machine-learning methods, investors can benefit from actively managed mutual funds.” In the second paper, researchers explore whether the use of artificial intelligence can help funds perform better. Researchers from the University of Macau evaluated the performance of artificial intelligence (AI)-powered mutual funds and found that these funds do not outperform the market per se. However, by comparison, AI-powered funds significantly outperformed their human-managed peer funds. They also showed that the outperformance of AI-powered funds is attributable to their lower transaction cost, superior stock-picking capability, and reduced behavioral biases. The first AI-powered public fund, AIEQ, was launched in 2017 and adopts machine learning technologies to actively select stocks in portfolio choices, according to the paper. The authors explained that AI-powered funds, in contrast to algorithmic trading, make decisions in the earlier stages of portfolio choices, use proprietary techniques to perform real-time prediction, and greatly enhance the flexibility and timeliness of traditional quantitative funds.