New property valuation technique delivers more accurate predictions

Researchers at the University of South Australia (UniSA) have developed a machine learning technique that makes property valuation more transparent, reliable, and practical, with the ability to accurately model the impact of urban development decisions on property prices.

How effective is the technique?

The technique was validated using over 30 years of sale information in metro Adelaide and uses purpose-developed machine learning algorithms to process data about housing, urban structure and amenities, to quantify the effects of urban planning policies on housing value.

Lead researcher, UniSA geospatial data analyst and urban planning expert Dr Ali Soltani, says the technique, has implications for the property, urban planning, and infrastructure sectors.

Dr Soltani says the model – developed in conjunction with Professor Chris Pettit from UNSW’s City Futures Research Centre – may also be extended to include other economic features at both the macro and micro levels, such as changes in interest rates, employment rates, and the influence of COVID-19, by harnessing the benefits of big data technologies.

What were the executive’s thoughts on the technique?

Commenting on the technique, Dr Soltani said, “Our modelling technique and findings may help real estate investors, builders, property owners, house appraisers, and other stakeholders gain a realistic view of the value of property and the factors that affect that.”

“This research has implications for policymakers by providing insights into the potential impacts of urban planning – such as infill regeneration, master-planned communities, gentrification, and population displacement – and infrastructure provision policies on the housing market and subsequent local and regional economy,” Dr Soltani further said.

“By capturing the complicated influence of infrastructure elements such as road and public transportation networks, commercial centres, and natural landscapes on home value, our model is especially valuable for enhancing the accuracy of current land value predictions and lowering the risks associated with traditional property valuation methodologies, which are largely dependent on human experience and limited data,” Dr Soltani commented.

“This model has the potential to be used as a decision-support platform for stakeholders, like home buyers and sellers, banks and financial agents, investors, the govt, and insurance or loan agents. Our technique makes it simpler for stakeholders and the general public to apply the findings of sophisticated models on historical or real-time data from multiple sources, which have previously been almost black-box and expert-oriented,” Dr Soltani concluded.