On the Potential of Multisensor Datafusion Approaches for Real Estate Valuation

Real estate valuation based on purchase prices is the approach closest to the market. However, in some submarkets only a few transactions take place and this approach fail because of the data. The usage of other real estate market data like offer prices, rents are obvious to use as extension of the data. To use this heterogeneous information, innovative methods for combination have to be derived. In multi-sensor data fusion first approaches already exist, e.g. for autonomous driving. In valuation context, offsets in data, different degrees of freedom (DOF) and accuracies play an important role. Reliable and secure decisions are made with the result of the fused data.In Multisensor-System information are measured in different scales and coordinate systems. Transformation is necessary to combine the data. The sensor accuracy (e.g. laser scanning vs. radar) and reliability (e.g. field of view) is very inhomogeneous and changing in different situations. These difficulties arise with real estate market data, too.In this paper we present a framework to adapt these fusion approaches for real estate market data. We discuss the used vocabulary in real estate context.The transferability of the term sensor is discussed in real estate market context. Basic topics of multi-sensor systems will be adapted like the structure of the system (competitive, complementary, cooperative) and the fusion levels (signal-, feature- or symbol-level). On signal level, only similar data can be fused, which observe the same phenomena. On feature level, there is a chance to merge derived data from different sensors, which have the same model variables (e.g. purchase prices and offer prices) and have the same DOF. On symbol level, data from different markets like rents and purchases can be combined.The derived framework can be used to make further investigations on market data in this context. The overall aim is the combination of different data with fusion approaches. It could be used in future to derive real estate market information in regions with few transactions.