Exploring new data sources to improve UK land parcel valuation

The paper describes a novel approach for building a UK-wide Automated Land Valuation Model and its implementation into commercial online software. We examine existing approaches to land valuation used in the UK, notably Trade Area Analysis, Spatial Interaction and Comparable Sales. We make the case that land use analysis, demographics and societal preferences affect the potential income and optimal use of parcels of land and hence the value of those parcels. This hypothesis leads to the introduction of a number of additional factors required to facilitate estimated land value, including traffic flow, population and site suitability. A number of artificial intelligence (AI) and machine learning spatial-temporal techniques are introduced to predict the value of all land parcels sold since 1995. We introduce a new technique, which includes (i) the application of Support Vector Machines to land use analysis; (ii) the use of predictive techniques for macro-environmental factors; (iii) the use of large, open-source data sets to improve valuation; (iv) industry alignment in predefined industrial tool. A number of different mathematical techniques are used to validate the proposed model and we show that our model demonstrates 92% accuracy for residential pricing predictions.