A Heuristic Robust Approach for Real Estate Valuation in Areas with Few Transactions

verfasst von
Alexander Dorndorf, Matthias Soot, Alexandra Weitkamp, Hamza Alkhatib

The German market transparency is mainly realizedby results of analyzingpurchase prices. Often, the purchases are analyzedin the context of aregression approach. The results are only reliable in areas with large numbers of purchases. However, in areas with only few transactions the solution of regression is not satisfactory. Furthermore, the purchase prices may contain outliers. Especially in areas with few transactions, the detection of outliers is a challenging task. This study presents three different estimation approaches which aredealingwith outliers. The first approach uses the data snooping to detect the outliers. The second approach is based on a heuristicRANSAC (random sample consensus) algorithm. The thirdapproach uses non–informative robust Bayesian regression techniques, in whichthe normal distribution of the likelihood data is replaced by a Student–distribution to ensure the robustness.The aim of this study is to investigate these three approachesin their efficiencyto deal with outliers in areas with few transactions. For this purpose a closed loop simulationiscarried. The results of the threerobust approachesare compared based on the knownregression coefficients and on the known observations. The results ofthe data snooping and RANSAC showthat the estimationfailmore often than the estimation by means of the robust Bayesian approach, which showsa suitable result forareas with few transactions.

Geodätisches Institut
Externe Organisation(en)
Technische Universität Dresden
Aufsatz in Konferenzband