WOZ valuation models

Municipality of Ede

Support in determining and verifying the WOZ value of residential properties.

Last change on 12th of December 2024, at 13:48 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

• Public finance
• Living

Begin date

2023-01

Contact information

[e-mail]

Link to publication website

https://www.ede.nl/aanvragen-en-regelen/woz-wet-waardering-onroerende-zaken

Responsible use

Goal and impact

The WOZ valuation models help determine the WOZ value of houses in the municipality of Ede every year. The algorithm helps appraisers check valuations fairly and consistently. The WOZ value is important for the level of property tax (OZB).

Considerations

The WOZ valuation models help valuers determine the WOZ values of homes properly and in the same way.

Human intervention

The WOZ value is officially determined by the levy officer according to Section 22 of the WOZ Act. According to article 17 of the WOZ Act, the value is determined. In practice, the WOZ assessor does this. The algorithm helps verify and support the valuations.

Risk management

The valuation function in the algorithm checks the values. The appraisers check the results by hand and adjust them if necessary.

Legal basis

Property Valuation Act (WOZ)

Links to legal bases

Wet Waardering Onroerende Zaken (WOZ): https://wetten.overheid.nl/BWBR0007119/2024-01-01

Operations

Data

The following data are used for the valuation:

1. Property characteristics: primary and secondary characteristics such as usable area, land area, year of construction, property type, state of repair, level of amenities, among others.
2. Market data: sales prices of comparable properties around the valuation date.
3. Location characteristics: distance to amenities, socio-economic context and physical environment

Technical design

AI/ML models are used to value properties based on a combination of statistical techniques and Hedonic Pricing Model. This model forms the basis of the analysis. The premise is that the value of a property is determined by the sum of the individual contributions of characteristics such as:

• Dwelling characteristics (e.g. use area, land area, year of construction)
• Location characteristics (e.g. proximity to amenities, neighbourhood demographics).
• Market condition (e.g. trends in supply and demand).
• Sales conditions

When valuing homes, the AI model uses LightGBM (decision trees). The algorithm analyses millions of combinations of characteristics (e.g. year of construction, type, usable area) and their influence on price. In doing so, it learns which combinations and patterns are statistically relevant. A decision tree divides the data into segments (e.g. homes of the same type or similar location) and determines which features have the most weight in predicting value. Visualisation is then made of which features contribute most to the final valuation.

The model is trained and validated in two steps:

• Training set (80% of the data): here, the model learns relationships between characteristics and transaction prices.
• Test set (20% of the data): the model is validated here to check whether it can also correctly value new, unknown data.

If the model performs well, it is applied to all properties. If not, it is re-trained and adjusted. To ensure accuracy, ratio checks are performed and the model is tested against international standards such as those of the IAAO.

External provider

Xxllnc