With the development of the digital economy, residents’ consumption behaviors have become increasingly diverse. This study employs big data analytics to explore the patterns of change in consumer behavior and the pathways of consumption upgrading. Building upon the traditional RFMT model, the study introduces the “T” indicator for recommendation traffic to construct an enhanced RFMT segmentation model. Additionally, the SOM neural network model is improved through optimized learning rate design to enhance training stability and clustering accuracy. Based on the consumption data of 4,158 households, clustering analysis identifies four typical consumer groups: core type, habitual type, supportive type, and general type. By incorporating five individual factors such as city tier, the study finds that the core type represents a growth engine composed of high-net-worth individuals, the habitual type reflects a pragmatic group with stable repurchase behaviors, the supportive type includes highpotential scenario-driven consumers, and the general type consists of price-sensitive long-tail users. Further analysis using eight indicators, including digital technology usage, identifies four consumption behavior types: technology-empowered consumption, interest-driven consumption, socially embedded consumption, and resourceconstrained consumption. The findings reveal that residents’ consumption behaviors are influenced by a combination of factors, with significant differences among consumer groups. The study recommends designing differentiated consumption upgrading strategies tailored to the needs of each group to expand the consumer market and promote high-quality development of the consumption economy.
