Development a Spatial Recommender System for Real Estate Marketing with Temporal Valuingcapability

The main goal of this research is to develop a knowledge-based recommendation system for real estate marketing. The system is capable of handling both spatial and attribute data of the properties. In general, there is no specific and stable relation between the parameters of the properties. In other words, we cannot extract any domain for the values of other parameters from the values of any given parameter. Therefore, the case-based method is used for the development of the recommendation system. The hedonic method is used for the property valuation and the geographical weighted regression method is used for extracting the coefficient of hedonic function. The valuation is based on the information of sold properties, considering different parameters such as construction year, number of building unit, frontage and distance from parks, schools, hospital, etc. In this system, information regarding for-sale properties are stored in the sale properties. Similarly the information regarding the preference and requirement of real estate buyers are stored in buyer profiles. Comparing this two, three of the most similar properties to the preferences of the buyer are recommended by the system. The system is tested using the information of sold properties in the “Vanak” region of district 3 of Tehran. Evaluation of the system, shows that the recommended properties are actually the most fitted ones to the requirement of the sample buyers.