Hedonic pricing models based on Machine Learning

This project will develop accurate estimates of real estate assets for both sales and rent operations that improve on the state of the art in industry. Properties include flats, chalets, offices and warehouses. In order to obtain real estate prediction values we use machine learning models, such as those based on Gradient Boosted Models. Input data used in this prediction process include:

– Supply side data: property characteristics such as constructed area, number of rooms, bathrooms, and other related characteristics such as the presence of swimming pool, air conditioning, orientation…);
– Spatial data: area features such as median real estate prices, accessibility measures, urban morphometrics, etc.
– Temporal aspects
– Demand side features: normalized visits, number of contacts, etc.

An important requirement of the models is that it is necessary to be able to easily explain point predictions. In this respect, the project will explore the use of model agnostic Machine Learning Interpretability (MLI) tools and other AI advances such as GAMs for this purpose.

Project reference: LV43a

Deadline 14th April