Lasso Penalty method for variable selection in database construction process and developing house value models in RUA

The aim of this paper is to confirm that in the case of the analysis of large data sets, the Lasso Penalty Method (LASSO) gives better results in the process of eliminating variables for the purpose of real estate value models than classical methods such as Ridge Regression. The selection of variables for an econometric model is closely related to its quality and suitability for intelligent decision support systems applied in the processes of real estate value management in the vicinity of airports. Airports face huge compensation payments as a result of the negative noise externalities they generate for neighbouring properties. Both the scale and complexity of this phenomenon require the development of intelligent systems to support airport decisions on compensation management and to monitor the environmental effects of these decisions. The selection of variables describing property characteristics is the first step in building a cost-effective system, as many of these characteristics are qualitative in nature. While systems can feed spatial information and data available in official registers, qualitative data often requires individual assessments and field visits. Hence, assessing the suitability of a given piece of information for a model is crucial at the data collection stage, especially when long time series are being constructed. For this purpose, it is proposed to use LASSO and then to model the value of developments consisting of detached houses on sets of variables selected with this method. The results obtained for LASSO are promising. They give the best set of qualitative and quantitative explanatory variables. The method has less variability than other subset selection methods tested. LASSO reduces some coefficients and zeros others, while retaining the positive attributes of subset selection and ridge regression. Furthermore, LASSO performs variable selection and coefficient estimation simultaneously. In the perspective of large set formation, this method of variable selection can also be used in Data Mining methods for estimating the value of large sets of properties. The obtained results can successfully support the process of property value management in the vicinity of airports.