Advanced forecasted valuation models for the real estate market (FutVal)

For most of us, a house is often the single largest investment we make during our lifetime. However, there are no tools available for predicting the future prices of your home. And with the Norwegian residential real estate market estimated to be valued at 9 trillion NOK - a significant fraction of the total economy - accurate and human interpretable models for future real estate prices are crucial for everyone, from individuals to analysts, investors, and public institutions. This project aims at developing a forecasting model by combining scientific advances in machine learning with more classical mathematical methods. The model must be applied in real-time with human interpretable outputs. We can use the forecasting model together with our existing estimates for the current real estate market values to provide historical, current, and future estimates of a home value on our real estate portal virdi.no. We hope that this can enable individuals, professionals, and non-professionals, to make more informed decisions in the real estate market.

On an individual level, a house is often the single most substantial investment a person makes during his or her lifetime, yet paradoxically there is no access to data-driven tools to assess forecasted values that could impact individual purchasing behavior. Moreover, on a societal level, the total Norwegian residential real estate market is an enormous asset class estimated to be an overwhelming 9 trillion NOK and is a significant aspect of the current economy and future business cycles. Sound techniques for modeling and forecasting the performance of real estate assets are, therefore, crucial for everyone, from individuals to analysts, investors, and public institutions. This project aims at developing an intelligent, real-time adaptive forecasting model by combining modern machine learning and classical mathematical methods, for forecasting future real estate prices. We aim to create a real-time forecasting model, which is still human interpretable, in contrast to a black-box approach. We can use the forecasting model in concert with our existing market-leading model for estimating the current real estate market values to provide a real-time current and future value estimation of the entire real estate market in Norway. The projections can be generated instantaneously and will be available for private parties on our real estate portal virdi.no, and a product we will sell to businesses. The goal is to help individuals and businesses to best manage our largest asset class, real estate.