The influence of discretization of the flat's area on its selling time

Many of the variables that describe property attributes are continuous. Additionally, during the digitization of the real estate market, analysts have access to a large amount of information. This overload of information (many observations, many attributes) makes it difficult to analyse the data and identify patterns. For continuous variables, therefore, discretization is recommended, which among other things: speeds up calculations, increases the transparency of results, facilitates interpretation, allows the use of variables with outliers.

The subject of the study is a mixed property offered for sale. The authors study the time of sale of properties depending on their area, using the method of duration analysis and the Kaplan-Meier estimator. However, these methods require attributes to be in categorical, nominal or symbolic form. In real estate research, continuous variables are most often discretized using expert knowledge and a lack of proposed solutions to this problem in the literature. There is also a lack of research on comparing the effectiveness of different discretization methods. Therefore, the paper proposes a discretization of real estate area according to thirteen selected methods. The result of the discretization should be, among others, disjoint intervals and significantly different time of sale in selected subgroups, which was the basis for assessing the effectiveness of selected methods.

Due to the large number of observations (3,732), methods referring to the number of observations and the range of values proved unsuitable, with equal-frequency methods generating a larger number of classes than equal-width methods. As a result of using methods discretizing the variable into too few classes (4), selling time did not differ significantly among subgroups. The best results were obtained using the minimum entropy method and expert indication of the interval span. The results can be used in the practice of selling residential real estate, as preliminary information about the time of sale depending on the area of the flat.