A0322
Title: A divide-and-conquer approach for spatiotemporal analysis of large house price data from Greater London
Authors: Soudeep Deb - Indian Institute of Management Bangalore (India) [presenting]
Kapil Gupta - Indian Institute of Management, Bangalore (India)
Abstract: Statistical research in real estate markets, particularly understanding spatiotemporal dynamics of house prices, has garnered attention in recent times. Albeit Bayesian methods are common in spatiotemporal modelling, standard Markov chain Monte Carlo (MCMC) techniques are usually slow for large datasets such as house price data. A divide-and-conquer spatiotemporal modeling approach is proposed to tackle this problem. The method involves partitioning the data into multiple subsets and utilizing an appropriate Gaussian process model for each subset in parallel. The results from each subset are then combined via the Wasserstein barycenter technique to obtain the global parameters for the original problem. The divide-and-conquer approach allows multiple observations per spatial and time unit, thereby offering added benefit for practitioners. As a real-life application, house price data of 0.65 million properties are analyzed from 983 middle-layer super-output areas in London for a period of eight years. The methodology renders insightful findings about the effects of various amenities, trend patterns, and the relationship of price to carbon emission. Further, as demonstrated from a cross-validation study, it records good predictive accuracy while balancing the computational need and is proved to be more effective than a conventional Bayesian approach.
