An AI-driven recommender system for housing-market predictions and decision making

When buying or selling a house a huge range of variables can affect your success. Your first impression starts with the estate agents listing, with many prospective buyers put off by poor quality photos or a lack of floorplan, let alone the price of the property.

This project was established with Strike to develop a recommender system which takes variant data streams (including customer and market data) and fuses them to recommend various actions based on accurate predicted outcomes of market trends. This will create efficiency savings for the agents and account managers and will provide a more personalised customer experience for their clients.

To begin with, we are exploring how we can use predictive analytics to determine the likelihood that a property may sell (and how fast) at specific pricing points. There are a multitude of factors at play: market conditions, the popularity of the property, the condition and hyper location of the property, the pricing of comparable nearby properties, the images on the property, and how the property is promoted, to name but a few.

In this project, customer and market data will be fused and mapped using knowledge graph ontologies. Additionally, stochastic forecasting and other methods will be used to analyse historical data to make insightful predictions.

Ultimately our recommender system will have three key impacts:

• Enable customers to take decisions which lead to faster, higher value house sales
• Allow Strike to effectively upsell financial services to existing clients
• Reduce the amount of human-customer interactions necessary to make a sale, creating efficiency savings.

Funding

This project is a Knowledge Transfer Partnership funded by Innovate UK.