Subscription and Redemption Prediction in Mutual Funds Using Machine Learning Techniques

There are various factors that derive mutual funds' flow. Among them, factors related to investment patterns and behaviors of the investors are highly informative. It is very beneficial for a fund manager to know who are the most probable investors that are going to subscribe to or redeem from a particular fund in near future. In addition, extracting the important underlying factors involved in this process helps fund managers to plan for optimizing their fund's performance. Our experiments on historic transaction data of about 400 mutual funds show that we can extract most informative patterns and use them to predict mutual funds' flow with relatively high accuracy. In addition, the proposed investors' ranking method gives a curated list for running more effective targeted campaign.