A large construction company wanted to know if it was possible to predict prices of new construction projects. Currently, this is a lot of manual work that also involves external brokers.
Building houses takes time and prices are very volatile in the current market. Prices can differ more than 10% from the start of a project to the finish. Accurate prediction is therefore very important for investors and construction companies. To be able to do that, you can use real estate agents to make a prediction. Trouble with that is that if you ask three different agents, you also get three different answers. This is because there are differences in market knowledge of each individual person. How can that be improved?
By analyzing large amounts of new houses and existing housing projects using machine learning, we have created an application that automates this work.
So we started gathering housing prices of all houses in the Netherlands. Houses that are currently for sale, but also very important those that are sold. Luckily there is a lot of publicly available information for every individual house.
Pricing of houses hugely differ from different neighborhood, even if they are comparable. Sometimes the street even makes a difference. It works best on cities because there is simply more data to compare.
Along the way the algorithm improves in its prediction. We used machine learning techniques to create a model for the prediction.