Cognitive computing for customer profiling: meta classification for gender prediction

Abstract

Analyzing data from micro blogs is an increasingly interesting option for enterprises to learn about customer sentiments, public opinion, or unsatisfied needs. A better understanding of the underlying customer profiles (considering e.g. gender or age) can substantially enhance the economic value of the customer intimacy provided by this type of analytics. In a design science approach, we draw on information processing theory and meta machine learning to propose an extendable, cognitive classifier that, for profiling purposes, integrates and combines various isolated base classifiers. We evaluate its feasibility and the performance via a technical experiment, its suitability in a real use case, and its utility via an expert workshop. Thus, we augment the body of knowledge by a cognitive method that enables the integration of existing, as well as emerging customer profiling classifiers for an improved overall prediction performance. Specifically, we contribute a concrete classifier to predict the gender of German-speaking Twitter users. We enable enterprises to reap information from micro blog data to develop customer intimacy and to tailor individual offerings for smarter services.