Combining Housing Price Forecasts Generated Separately by Hedonic and Artificial Neural Network Models

Aims: A) To enhance accuracy in forecasting housing unit prices by forming combinations of component forecasts generated separately hedonic and artificial neural network models; B) help ascertain whether a constrained or unconstrained linear combining model achieves superior performance.
 Place Duration the Study: Department Business Administration, Istanbul Aydin University, 34295, Turkey; from 2019 to 2020.
 Study Design: A cross sectional data set corresponding attributes characteristics is formed then randomly divided into two segments: sample (80%) out (20%). Three different methods (hedonic, combining) are employed process same set, generate forecasts. The three tested compared.
 Methodology: Out combination with forecast weights weighted least squares (WLS) regression realized price against Four types regressions run: unconstrained, without constant; constrained, constant. Then mean absolute error each method calculated difference between all pairs models compared nonparametric Wilcoxon sign rank test.
 Results: constant term suppressed sum-of-the-coefficients equal one, generally performs best, comparison other (component combination) examined study.
 Conclusion: findings represent further evidence regarding benefits applying constraints on model; demonstrate that can be successful technique for enhancing prices.