Home|Blog|Determining the VON price with AI

June 18, 2021

Determining the VON price with AI

VON prijs bepalen
2.3 min readBy Published On: 18 June 2021

You have a real estate project in mind and it is time to make an offer. But how do you know whether you're bidding too little or - even worse - too much? By using Artificial Intelligence (AI) you determine the optimal real estate price and you gain insight into why it is that price. So you no longer have to make decisions based on gut feeling. In this blog we explain how to determine the VON price with AI.

When determining the VON price with AI, the first thing you need is data. We want to include all factors that influence the average house price. Our application Houzr uses data about real estate projects from recent years. We supplement this with data from Statistics Netherlands, for example, about demographics, WOZ values, incomes, employment, but also the proximity of all kinds of facilities (think of shops, restaurants, greenery) and accessibility, noise pollution, etcetera.

Determining the VON price with a data model

We then use a model called XGBoost Regression along with transfer learning. XGBoost Regression is based on decision trees. With a decision tree you start with a house. Suppose the surface is larger than 100 m2, then you end up in a bucket with all houses of that surface. Then, for example, you look at how big the garden is and you end up in yet another bucket. You get deeper and deeper into the decision tree, until you get to the last bucket and then voila: there’s your prediction. An XGBoost regressor combines multiple decision trees in an advanced way, giving us an even more reliable prediction.

We know that data on existing buildings could also help predict VON prices for new homes. Yet the properties of existing buildings and new buildings are too different to simply merge these data. That is why we also use Transfer Learning (TL) in addition to XGBoost Regression. This method ensures that we can still add the extra information from existing buildings to our model, without it being at the expense of the information we get from the new-build homes.

Understanding outcomes 

XGBoost is seen as a so-called Black Box AI: you put something in it and something rolls out, without knowing why. We actually don’t want that, because algorithms that take on a life of their own can go in the wrong direction (as happened at the Tax and Customs Administration, for example). That is why we use SHAP Values, a method that we overlay our model to gain more insight into it. We want it to be clear which variables influence the ultimately determined VON price. For example, it is possible to see that the cafe around the corner makes the house worth 50 euros more per square meter.

Set your own VON price with AI?

With Houzr you as a project developer can calculate and test large quantities of possible project configurations at possible construction sites in the Netherlands. The Houzr algorithm quickly calculates hundreds of thousands of scenarios and uses various data sources and prediction models to find the optimal balance within the set frameworks between your return, risk and market fit.

Curious about the optimal bid for the item you have in mind? Request a free demo account and start optimizing.

Share this post with your network 👉🏼

Wanna read more?

Get a free tactical attack plan.

Call us today at(010) 4047 333

Free consult

Don't worry: it's completely non-binding.