A Method of Analysis on Consumer Behavior Characteristics Based on Self-supervised Learning

Abstract

Consumer shopping decision-making style is a kind of mental orientation, which represents the way consumers make decisions and has cognitive and emotional characteristics. It determines the behaviors of consumers and is relatively stable for a long time, as a result of which can be used as the basis for market segmentation. The traditional way to determine shopping decision-making style is mostly through questionnaire, which is time-consuming and laborious. This paper uses e-commerce data, combines the different commodities purchased by consumers, and comprehensively measures the mental characteristics of consumers when they make purchasing decisions based on their behaviors. By clustering analysis, the shopping decision-making behaviors of consumers are divided into two categories. Considering that clustering is an unsupervised method, its results are often not perfect, which is especially reflected in the data at the junction of two categories. In view of this, we divide the data at the junction of two categories into unclassified data and then train naive Bayes with classified data to classify unclassified data. Ultimately, by synthesizing all shopping decision-making behaviors of each consumer, the decision-making styles of consumers are divided into three categories: direct style, cautious style and neutral style. The experimental results show that the model proposed in this paper makes the classification of consumer decision-making behaviors more intuitive and is obviously superior to the comparison model. This model can effectively determine the consumer decision-making style.