A deep learning approach for estimation of price determinants

Abstract

This paper aims to investigate the spatial price dispersion and estimate the price determinants of a homogeneous product sold by marketing channels in the industrial automation market in the south of Germany. Empirical studies reveal that different prices are charged for the same homogeneous products, which shows a clear deviation from the law of one price. Understanding the key determinants of the prices charged by the marketing channels enables manufacturers to design appropriate channel selection criteria with regional differentiation and to improve their existing channel performance management. This research proposes a new geomarketing approach that combines spatial and hedonic price modeling by using deep learning. Since ANN models are often perceived as a “black box”, this paper demonstrates a technique for extracting knowledge from an ANN to leverage the market intelligence. This study reveals the causal relationships between the explanatory variables and the price within the ANN model. The results indicate that deep learning models provide higher accuracy in modeling the price and its determinants compared to statistical models. However, the econometric model GWR enables detecting the geographical variation, also called spatial heterogeneity, among the retained variables to explain the patterns of price, especially their volatility across space.