A few words about the algorithms
The algorithms used vary and are different for every data set used. They can be very simple such as the hedonic regression in different forms (including the GWR – geographically weighted regression – a combination of hedonic regression with the use of GIS). A Hedonic regression-based model, breaks down a property's value into its individual components, such as area, number of bedrooms, location, age, floor, state of repair and other features. The model then assigns weights to each component and uses those weights to predict the value of a property based on its characteristics. However, properties as commodities are not homogenous and even neighboring pieces of land can have great variations in value. This heterogeneity of properties most of the time prevents those algorithms based on hedonic regression from showing great accuracy. The statistical measures for the successful implementation of such models can be the R2 (coefficient of determination), the MAPE (mean average percentage error) and the COD (coefficient of dispersion-the average difference a group of numbers has from the median).
While Hedonic based regression algorithms fail to pass the statistical tests mentioned above, other models based on Machine Learning and neural networks show much better results. This type of algorithm can learn from and make predictions on complex data sets. In this case, the neural network would be trained on historical sales data and property characteristics to predict the value of a property under study. Similar models and algorithms are based on decision trees theory (random forests for example), k- nearest neighbors, support vector machines and many others.
Bibliography and journal publications indicate that AI and Machine Learning based Models are offering higher accuracy but based on my personal experience after using them for several years, they have two common problems, the “generalization” and the “overtraining”. Generalization is the ability of a machine learning model to perform well on data that it has never seen before. When a model is not able to generalize well, it means that it has become too specific to the training data and may not perform well on new or unseen data. This is also known as underfitting. The model may not capture all of the patterns and relationships in the training data, resulting in poor performance on new data. Overtraining, also known as overfitting, occurs when a model is trained too much on a specific set of data, and as a result, it becomes too specific to that data. This means that the model may perform very well on the training data, but it will not generalize well to new or unseen data. Overtraining occurs when a model is too complex, and it can capture the noise or random variations in the training data, rather than just the underlying patterns and relationships. This can lead to poor performance on new data.
Can AVMs replace traditional valuations?
In my personal view as a person who really enjoys dealing with mathematics, I believe that AVMs can replace partially traditional valuations but we are not ready yet for that transition. Currently, they can be just useful tools for an indication of the value of the property, but they cannot entirely replace traditional valuations conducted by professional appraisers. We should also have in mind that the valuation profession is based on an expert’s opinion which is formed on his experience and the on-site inspection in order to get the unusual and unique features of the property that are not normally captured in the available information.
Such models can be fully implemented successfully only when we have sufficient data that can be produced from sensors installed and other mechanisms. Until then, traditional valuations will be preferred by the majority of the clients who are willing to pay a lot for an expert’s opinion.
It's important to note that AVMs are not a substitute for a professional appraisal yet, and may not be appropriate for all types of properties or valuation purposes.
Who is using AVMs and what are CAMAs?
AVMs are often used by lenders, real estate agents, and appraisers to quickly estimate the value of a property. A very common form of an AVM is the CAMA (Computer Aided Mass Appraisals). CAMA systems are often used by local governments to assess property taxes. They allow assessors to value large numbers of properties quickly and with relevant accuracy which can save time and resources compared to traditional manual appraisal methods. CAMA systems can also provide more transparency in the valuation process and reduce the potential for bias or inconsistency in property valuations.
In the decision-making process to whether a Mass Appraisal System shall be used, the outermost important factors that I would recommend are:
• Initial cost
• Available data.
Every single property valuation is a unique project and has a clear start and end date. Manual valuations are usually both time and money resource intensive and often deliver results in crucial revaluations later or sometimes never. In a project, there is always a trade-off between Time, Money and Quality. Increasing one of the factors almost automatically decreases the remaining two. For example, a valuer who tries to complete more valuations within a given period, either must decrease the quality of each valuation to be faster per valuation or must hire more staff to deliver more valuations. AI does not have any of these constraints. It can work 24/7 and with the correct data and can produce a theoretically infinite number of valuations. Practically, the amount is limited to the available data as well as the input of this data by a human source.
Data, resources and cost.
Data are the most crucial component. CAMA and AVM can only exhibit high computational efficiency if the database contains adequate data. Theoretically, one could state that if no data is available, AI cannot be used. On the other hand, without precise data, any human-based valuation would not be very precise either. It takes years of studying and obtaining practical experience as well as local market knowledge for a valuer to be able to deliver accurate valuations and appraisals. This process of learning is time-consuming and rather expensive. AI can do so within a short period of time and can improve its performance based on past observations. Due to this, human valuers are considered to be expensive. AI can offer a much less expensive rate for any valuation since costs such as travel time and travel expenses to the property can be saved. However, AI has a higher initial cost, as it is expensive to set up a model. The maintenance of the database and feeding the AI model with more data are usually the highest running expenses.
Responsibility, professional liability and technological process
Many scientists stated that feelings and sympathy are what make us human. These are unarguably great assets of every human; however, in valuations, they can create inaccuracies due to the loss of neutrality. Humans can only control their doings up to a certain level. AI does not lose neutrality and hence accuracy, due to sympathy, therefore, in this aspect it can create more accurate valuations. Carrying out an official valuation requires, in almost every country, a license. These licenses are often provided by human-based associations. Often political reasons block any technological process as some humans fear losing their job to AI. This political lobbying reduces progress considerably and by doing so the human valuer is heavily favored. Human valuers often argue about the responsibility and legal pursuit of AI. A valuation carried out by a human valuer can always be challenged and one can sue the person who completed the valuation. But, the questions to be answered are who do you sue when an automated Mass Appraisal valuation is in question, and who signs an automated Mass Appraisal valuation. The above two questions can unfortunately not be answered easily. Looking for the responsible party of a Mass Appraisal valuation is a tricky process, which is one of the major drawbacks of AI. However, if we feed the AI model with enough data and constantly maintain and update the database, the possible margin of error shall be small enough to be negligible, and costly legal processes could be avoided or minimized.
The article was prepared by Dr. Dimopoulos Thomas for AaRVF (Assessors and Registered Valuers Foundation, India - https://aarvf.org/)
You can read the full article here