Outlier Detection for Analysis of Real Estate Price

Abstract

The original form of the real estate data, which is used for different purposes, such as housing or investment, needs to be preprocessed for outlier detection caused by spelling errors, systemic errors or artificial documents created to attract attention. In Literature review, there is no analysis study of the real estate price which uses multi-variate outlier detection. In this study, different outlier detection methods have been conducted and the results are compared using the housing prices in Istanbul province by considering the register office information. The models were tested on the labeled data-set which is located in Istanbul. The best results were obtained by using One-class Support Vector Machines (OCSVM) and Average K-Nearest Neighbor (AveKNN). After elimination of the detected outlier documents, a general system is created to generate the ranges for price per square meter of the real estate values. The best results to estimate the range for price per square meter were obtained using Average KNN model. The subject of this study, determination of price per square meter ranges of real estate values are expected to be used in various areas, such as investment, tax auditing, detection of false advertisements on real estate sites.