Authors: Anton Gerunov
Addresses: Sofia University 'St. Kliment Ohridski', Bulgaria
Abstract: This paper utilises a novel experimental dataset on consumer choice to investigate and benchmark the performance of alternative statistical models under conditions of extreme uncertainty. We compare the results of logistic regression, discriminant analysis, naïve Bayes classifier, neural network, support vector machine (SVM), decision tree, and random forest (RF) to discover that the RF model robustly registers the highest classification accuracy. Variable importance analysis reveals that apart from demographic and situational factors, consumer choice is highly dependent on social network effects and subject to numerous non-linearities, thus making machine learning approaches the best modelling choice.
Keywords: choice; decision-making; machine learning; uncertainty; social network; logistic regression; neural network; random forest; consumer choice; modelling.
International Journal of Business Intelligence and Data Mining, 2019 Vol.14 No.1/2, pp.238 - 253
Received: 18 Jan 2017
Accepted: 03 Feb 2017
Published online: 16 Nov 2018 *