Enhancing Intelligent Property Valuation Models by Merging Similar Cadastral Regions of a Municipality

摘要

A method for enhancing property valuation models consists in determining zones of an urban municipality in which the prices of residential premises change similarly over time. Such similar zones are then merged into bigger areas embracing greater number of sales transactions which constitute a more reliable basis to construct accurate property valuation models. This is especially important when machine learning algorithms are employed do create prediction models. In this paper we present our further investigation of the method using the cadastral regions of a city as zones for merging. A series of evaluation experiments was conducted using real-world data comprising the records of sales and purchase transactions of residential premises accomplished in a Polish urban municipality. Six machine learning algorithms available in the WEKA data mining system were employed to generate property valuation models. The study showed that the prediction models created over the merged cadastral regions outperformed in terms of accuracy the models based on initial component regions.