The simulated consumer – An agent-based approach to shopping behaviour

Abstract

Recent research on consumption has indicated a growing differentiation of consumption styles and habits, which in turn make the shoppers' decisions increasingly hard to grasp and more complex in general. In particular, it is found that the classic modelling techniques fail to sufficiently represent these decision processes. This paper examines the potentials of an agent-based approach for the locational choices of consumers for their shopping activities on a regional level. The model is applied to two product sectors, groceries and clothing, in two distinct research areas in northern Sweden and southern Germany. While the empirical findings generated for the two case studies resulted in quite different model designs, both are found to be equally applicable for representing the different choice mechanisms and result in forecasts of buying power flows and thus turnovers of individual shops. It is concluded that agent-based approaches provide an elaborate instrument to model consumer locational choices and to forecast consequences of planning decisions, although expanding the empirical basis on qualitative grounds of research may lead to still further advancements. Zusammenfassung: Die Forschungen der letzten Jahre zeigte, dass Konsumstile und-gewohnheiten sich immer weiter ausdifferenzieren. Dies hat zur Konsequenz, dass Entscheidungen der Verbraucher immer schwerer für die Forschung und Modellierung fassbar werden. Insbesondere ist festzustellen, dass die klassischen Modellierungstechniken bei der Darstellung dieser Entscheidungsprozesse an ihre Grenzen stoßen. Der vorliegende Beitrag untersucht die Potenziale eines agentenbasierten Ansatzes zur Modellierung der Einkaufsstättenwahl auf einer regionalen Ebene. Das Modell wird in zwei verschiedenen Untersuchungsräumen (in Nordschweden und im südlichen Deutschland) auf zwei Produktbereiche, Lebensmittel und Kleidung, angewendet. Obwohl die empirischen Befunde zu den beiden Fallstudien zu verschiedenen Modell-Designs führten, können beide Modelle gleichermaßen eingesetzt werden, um nicht nur die Einkaufsstättenwahl von Individuen darzustellen, sondern auch Abschätzungen von Kaufkraftströmen und Umsatzschätzungen einzelner Geschäfte vorzunehmen. Die Ergebnisse zeigen auch, dass agentenbasierte Ansätze ein sehr gut geeignetes Instrumentarium bieten, um nicht nur das Handeln und Verhalten von Akteuren zu modellieren, sondern auch um Folgen von Planungsentscheidungen zu simulieren. Potenziale bestehen jedoch noch in der Einbindung qualitativer empirischer Ergebnisse in die Modelle.

Simulating Shopper Behavior using Fuzzy Logic in Shopping Center Simulation

“…Literature study We used a large number of books, journals, reports, and websites relating to the research topic, especially about the multi-agent system, artificial intelligence and shopping center simulation. References [2][3][4][5][6][7][8][9][10] were used to provide the basis and rules in building the simulation tool, while other references were used to assume and extract the attributes that the agents can possess.…”

“…The fuzzy rules were built and referenced from literature taken from social science [2][3][4][5][6][7][8][9][10] and from experts' recommendations. As for attributes such as income or age, they are fairly easy to model by referencing official documentation in each country.…”

“…Consumers and tenants are treated as individual agents with their own thinking capabilities and preferences [9]. Environments such as roads, floors, and walls are treated as passive objects, which are mainly used as pathfinding variables.…”

“…It basically lowers the constraint of the fuzzy rules and gives shopper agents an unpredictable and more complex behavior towards their preferred activities, as shown in Figure 9. The fuzzy rules and membership functions were constructed and obtained from various publications on social multi-agent systems [2][3][4][5][6][7][8][9][10]. The rules are stringbased and are parsed within the application.…”

“…In the game domain, fuzzy logic not only acts as an AI method but also as a quality behavior generator, which is able to provide complex behavior and increase the quality of interaction between agents [1]. The knowledge base and rules were taken from various publications on consumer behavior and analysis [2][3][4][5][6][7][8][9][10]. The design of the agent itself contains its knowledge base and decision rules crafted in fuzzy logic, but the agents have very little knowledge of the virtual environment and the other agents.…”