Predictive modelling of future tea consumption: analysing consumer behaviour with combining the machine learning and agent-based simulation approaches

Abstract

The study aims to develop an agent-based simulation model combined with a machine learning method to predict changes in consumers’ tea consumption habits and perceptions due to various factors and the industry’s status. It was carried out specifically from Turkey, the world’s leading tea consumer. The data were collected via survey based on the whole country. The model uses machine learning technique to analyse brand and per capita tea consumption data, and rules are generated for these variables. The model, which was verified and validated, examined the structure of tea consumption in 2050 and 2075. The results showed that tea consumption per capita will increase compared to that in the current situation. In a scenario where tea prices increase because of agricultural and climate factors, there is no direct decrease in demand, suggesting that action should be taken for the sustainability of tea before facing these kinds of situations.