Fund2Vec: Mutual Funds Similarity Using Graph Learning
Identifying similar mutual funds (including exchange-traded funds) with respect to the underlying portfolios has found many applications in fund recommender systems, competitors analysis, marketing and sales of the products. The traditional methods are either qualitative, and hence prune to biases and often not reproducible,or, are known not to capture all the nuances (non-linearities) among the portfolios from the raw data. We propose a radically new approach based on the weighted bipartite network representation of funds and their underlying assets data using a sophisticated machine learning method called Node2Vec to learn an embedded low-dimensional representation of the network. We use this network embedding to identify similar portfolios by computing node similarities in the representation space, which we call Fund2Vec. Our approach provides novel insights to the portfolio similarity problem as well as a data-driven method to remove bias from qualitative categorizations available in the market. Ours is also the first ever study of the weighted bipartite network representation of the funds-assets network.
Key Takeaways: 1) Machine Learning for mutual funds analytics (sales, marketing, portfolio diversification, etc.) 2) An interesting usecase for application of graph machine learning in investment management industry 3) A bit technical details on a specific graph machine learning algorithm called Node2Vec
Dr Dhagash Mehta is a Senior Investment Strategies Manager (Machine Learning and Asset Allocation) at Investment Strategies Group at Vanguard, and prior to that was a Principal Research Data Scientist at Vanguard. Dr Mehta is an Editorial Board Member at the Journal of Financial Data Science (https://jfds.pm-research.com/). Dr. Mehta pursued his undergraduate studies in Physics in India, followed by Part III of Mathematical Tripos at the University of Cambridge, and Ph.D. in theoretical particle physics from the University of Adelaide (Australia) as well as Imperial College London (UK). Before joining Vanguard, he was a Senior Research Scientist at United Technologies Research Center (now called Raytheon Technology Research Center), and prior to that a Research Professor at Department of Applied and Computational Mathematics and Statistics at University of Notre Dame. He has held multiple research positions at various research institutes such as Fields Institute in Toronto, Simons Institute for Theory of Computing at Berkeley, the University of Cambridge (UK), Imperial College London (UK), the University of Adelaide (Australia), North Carolina State University (USA), Syracuse University (USA) and National University of Ireland Maynooth (Ireland). Dr. Mehta’s areas of expertise are theory of machine/deep learning, and applications of machine learning in finance.. In particular, he has published 75+ research papers in reputed journals on optimization (convex and nonconvex), computational algebraic geometry, numerical analysis, network science and machine learning to solve various problems arising in financial services and wealth/asset management (and in the past, power systems and control theory; theoretical physics, jet-engines, and smart building systems).