Dr. Sumeet Dua

Max P. and Robbie L. Watson Eminent Scholar Chair

  • Full Screen
  • Wide Screen
  • Narrow Screen
  • Increase font size
  • Default font size
  • Decrease font size

Preet K. Sekhon (2009)

E-mail Print PDF

Automated Valuation Models Using Data Mining Techniques- An Application to Real Estate Valuation

The most widely used approach used in valuation of Single Family Residential homes is sales comparison approach, where the subject is compared to most similar properties sold in the subject neighborhood. AVM’s (Automated Valuation Models) is basically a computer program built around statistical-mathematical analysis of comparable sales from prior sales transaction database. Almost all AVM’s are based on hedonic price theory, where each property can be defined by set of characteristic attributes. The main problem to develop AVM basically deals with identifying these attributes, assigning significance to these attributes and developing regression models on these attributes.

This project proposes using Minimum Spanning Tree based clustering to identify neighborhoods in city of Turlock and then develop valuation models on these neighborhoods. Three approaches are used to develop regression models. In first approach models are built on all the previous 7-8 months sales without any consideration to their location or important characteristics. In second approach the sales are clustered using their physical distance from each other and models are built on these clusters. The premise is that models built on spatial information should predict better as the variation in price due to location factors is minimized. In third approach, the clusters are formed using three significant attributes that may define submarkets without being spatially connected. Each test case is run through different models to test the predictive power of models. Models are compared using Mean Absolute Percentage Error (MAPE) and correlation between predicted and actual values.

You are here: Research Student Thesis