Back to results list
Please use this identifier to cite or link to this item:
|Title:||Towards property valuation accuracy : a comparison of hedonic pricing model and artifiical neural network||Authors:||Abidoye, Rotimi Boluwatife||Advisors:||Chan, P. C. Albert (BRE)||Keywords:||Real property -- Valuation
Real property -- Valuation -- Nigeria
|Issue Date:||2017||Publisher:||The Hong Kong Polytechnic University||Abstract:||The need for accurate property valuation in any country cannot be underestimated due to the significant relationship between the real estate industry and the national economy. Investment decisions relating to the acquisition or disposal of real estate assets are largely dependent on valuation estimates. Inaccuracies in property valuation is a global problem which have been of interest to all stakeholders, and Nigeria is no exception. A possible explanation for this might be the valuation techniques currently adopted, with the hedonic pricing model (HPM) being an example. As evidenced in previous literature, the HPM approach has gained wide acceptance among real estate researchers, despite its shortcomings. There is an obvious need to seek innovative approaches to improve the quality of property valuation estimates. To address the shortcomings of the HPM approach in property valuation, the current study sets out to develop prediction models for property valuation in Nigeria. Two modeling techniques, i.e. HPM and artificial neural network (ANN), were applied in this study. The predictive accuracy of the developed models served as a basis for comparison. The research objectives are to; assess the current property valuation practice in the Lagos metropolis; identify and generate a list of attributes that influence residential property values in the Lagos metropolis property market; develop a hedonic pricing model for the Lagos metropolis residential property market; develop an artificial neural network model for the Lagos metropolis residential property market; evaluate the predictive accuracy of HPM and the ANN model developed for the Lagos metropolis residential property market; and assess the Nigerian valuers' receptiveness to the application of artificial intelligence (AI) techniques in property valuation. The data used for this study were sourced from registered real estate firms operating in the Lagos metropolis, Nigeria. The information collected includes the awareness and adoption of various valuation methods, residential property value determinants and the receptiveness of valuers to AI techniques in property valuation, and they were gathered via the administration of online-based questionnaires to the valuers. The data were analyzed using the mean score (MS) ranking technique and the chi-square test. In addition, the sale transactions information of residential properties situated in the Lagos Island property market were collected. These were fitted into HPM and the ANN model developed for the Lagos metropolis property market.
The present research found that the valuers are aware of and adopt the traditional valuation methods, especially the comparable, investment and cost methods in practice. Whereas there is a little awareness and non-adoption of the advanced valuation methods. This suggests that there is a need for a paradigm shift towards more accurate and reliable property valuation approaches. The analysis of the property value determinants reveals that the structural attributes are the most significant to property value formation in the Lagos metropolis. This set of attributes were used for the modeling of the property values in the study area. In order to facilitate a justifiable comparison, the HPM and the ANN models were developed with the same data set. This data set was divided into two parts; for the training and the testing of the developed models. The HPM approach generated a coefficient of determination (r2) value of 0.77, but this did not translate into the prediction of accurate property values because of the high mean absolute percentage error (MAPE) value of 38.23% recorded. This is coupled with high mean absolute error (MAE) and root mean squared error (RMSE) values recorded as well. The ANN model produced an r2 value of 0.81 and a MAPE value of 15.94%. These values together with lower MAE and RMSE values are more encouraging when compared with these of the HPM approach. This indicates that the ANN model is a better substitute to the HPM approach in property valuation. Also, a large percentage of the ANN predicted property values generated prediction errors which are within acceptable international standards, when compared with the HPM outputs. The investigation into the receptiveness of Nigerian valuers to AI valuation techniques shows that the majority of the valuers are willing to acquire the competence in the application of the AI techniques in property valuation. When other requirements of developing a robust property valuation model are met, coupled with this high level of willingness of the valuers, the prevalence of property valuation inaccuracy in the Nigerian context could be reduced remarkably. Overall, the Nigerian property valuation practice is still at a traditional level. Also, structural attributes of properties were found to be the most important attributes affecting its value. The developed ANN model provides a tool which can be used for property valuation. In addition, the findings provide evidence which justifies the need to adopt advanced modeling techniques (such as AI technique) in the property valuation research and practice.
|Description:||xx, 300 pages : color illustrations
PolyU Library Call No.: [THS] LG51 .H577P BRE 2017 Abidoye
|URI:||http://hdl.handle.net/10397/70359||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|991021965755503411_link.htm||For PolyU Users||167 B||HTML||View/Open|
|991021965755503411_pira.pdf||For All Users (Non-printable)||3.18 MB||Adobe PDF||View/Open|
Citations as of Sep 24, 2018
Citations as of Sep 24, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.