THE chattering classes could soon be turning up for dinner parties clutching
their laptops, if software that can accurately forecast house prices鈥攖hat
perennial conversation topic鈥攖akes off. A prototype program based on
artificial-intelligence routines has already proved itself accurate to within
3.8 percent when predicting house price movements in Britain over the last three
years.
The AI-based software package is being perfected by Ian Wilson鈥檚 team at
Glamorgan University, with funding from the British government鈥檚 Engineering and
Physical Sciences Research Council. 鈥淲e want to provide more informed, objective
opinion about how property values might move and what the consequences would be
for a particular consumer,鈥� Wilson says.
In his lab, they are teaching different types of neural networks the factors
that influence house prices by feeding them government statistics going back
more than 20 years. The neural network that predicts most accurately will be
used in the final package. It will help house-hunters work out what鈥檚 likely to
happen if they buy a property at a particular price at a certain time in the
economic cycle.
Advertisement
Initially, the team will produce national house-price predictions, but it
plans to move on to regional and then local forecasts, a mid-December AI meeting
in Cambridge will hear. 鈥淚t will be very interesting to see how statistical
forecasting compares with AI-based forecasting,鈥� says Wilson. And it will be
even more interesting to see whether laptops will be flourished after the
蝉辞耻蹿蹿濒茅.