The Decade of the Data Translator The Decade of the Data Translator
Get your ticket or log in to build your agenda.

A Home Recommender System to Promote Path to Homeownership


Anshul Agarwal, PhD
Home Partners of America, Data Science Director

Anshul Agarwal is an industry expert with a decade long career in the field of data science, machine learning, computational methods, and operations research.  He has worked across several industries ranging from chemical, oil & gas, supply chain, airline, revenue management, retail, residential real estate and finance.  He is currently leading the data science division at Home Partners of America.  He has written numerous publications and maintains a hand-on approach generating innovative ideas in his field of expertise.  He holds a PhD from Carnegie Mellon University focused in Operations Research.


Home Partners of America (HPA) helps families access high-quality single-family rental homes in desirable communities and provides them with a potential path to homeownership through its unique resident-led Lease Purchase (LP) Program. HPA acquires single-family homes exclusively through this program which are chosen by applicants themselves from market listings, and leases are executed to rent. Promoting awareness of such a program to renters can bring them closer and sooner to owning their dream home. In this talk we describe a unique data-driven home recommender system (HRS) to expand the reach of the LP program. Based strictly on a user’s home preferences and desired features in a home, the HRS not only recommends potential rental homes to a renter, but also introduces the LP program while recommending active listings they could consider renting and owning. The model particularly ensures fair housing laws are followed and there’s no geographic or other kinds of discrimination in recommendations. 

In this talk, we describe the technical details of the model as well as design considerations specific to this problem. Because of the nature of user-home interactions, the HRS is built on a novel unique mixed-integer linear programming (MILP) framework. The talk will cover why traditional algorithms such as collaborative filtering or matrix factorization are not pertinent, and how the MILP framework addressed the needs appropriately.