Exploring the prediction of variety-seeking behavior

As a common choice strategy for consumers, variety-seeking has a direct impact on the business performance of enterprises, and has been studied for decades in management science and marketing research. Existing research tends to find factors which influence variety-seeking by means of questionnaires and laboratory experiments. Based on the important factors, marketers develop marketing plans which are tailored to consumers. However, due to personal motivation, the results of questionnaires and laboratory experiments may not fully reflect internal states of participants. Thus, we propose a data-driven framework, using ensemble learning for predicting variety-seeking behavior based on real student consumption data. Experiments demonstrate that our model outperforms separate machine learning algorithms and can effectively predict variety-seeking behavior. We further analyze the contribution of each feature towards the prediction, achieving some useful conclusions for the study of variety-seeking intervention mechanism.