There currently exist informatics applications – some of which may be consulted on the Internet – that offer a valuation of a property by means of a multivariate regression statistical process, including the surface area in square metres of the dwelling, if it has a lift, the location, if alterations or refurbishments have taken place, selling prices over the past few years, and so on. These systems are based on statistical values which are annually updated.

The project by the research team at the Public University of Navarre is more ambitious: it aims to get results closer to the market values of the properties than to those values supplied by the statistical method currently used. The objective is that the informatics application take into account, apart from the statistical aspects mentioned, the subjective valuation that a property appraiser might make.

Learning from the appraisers

This is why the experience and of the professionals or expert knowledge need to be introduced into the computer programme. And this is carried out by means of artificial intelligence, using learning techniques based on real cases.

Thus, the computer is taught the meaning of expressions like, “the house is a disaster” which, for some valuers, could mean that “the original construction was very bad, the orientation is awful, the distribution of the rooms impossible or it is a ruin”, while for another appraiser it might mean that, “the dwelling is far from the centre, it is on a troublesome estate, and so on, etc.”. The objective is for the system to learn the meaning of, “extrapolating the average of what each of the valuers means ”.

This process of intelligent learning is based on “neuronal networks”, so called because they try to imitate the functioning of the neurones in the human brain. The networks receive incoming information (the statistical characteristics of the dwellings) on the one hand and, on the other, information on how the minds of valuers work. These data are subject to a number of operations at different levels and, finally, the system comes up with an approximate valuation for the property.

However, brain neurones also react to stimuli but, given that we are at the dawn of history of artificial intelligence, for the moment all the operations have to be carried out from a mathematical viewpoint.

Artificial intelligence can be understood in an overall way, taking in everything: vision, natural language, movement, and so on. This involves trying to imitate the human in everything, including its creativity and emotional intelligence and, in this sense, artificial intelligence is still “very poorly developed”. This said, if artificial intelligence is understood as being that tool that is going to solve a specific problem, such as the dishwasher or washing machine that calculates the amount of detergent needed as a function of dirt, “then it works well”.

Intelligent zoning

In the market currently, the greatest conditioning factor in the value of a property is the area in which it is located. This zoning factor is currently assessed “by hand”. However, the location of streets and estates is not the only factor to be considered but others have to be taken into account. This is why the next step for the research team at the Public University of Navarre is to design a system enabling the intelligent zoning, by house prices, of a city, a province or a region.

Note: If no author is given, the source is cited instead.

• Artificial Intelligence
• Statistics
• Robotics
• Communications
• Computer Science
• Computer Programming

• Computing power everywhere
• Data mining
• Mathematical model
• Experimental economics
• Computer vision
• Speech recognition