A Cloud Computing-Based Framework for Enterprise Marketing Data Analysis and Customer Behavior Prediction in the Digital Economy Era

Abstract

This paper first establishes a data collection and quality control mechanism, combining statistical analysis tools and data mining techniques to enhance data value. It defines the boundaries of customer behavior, introduces survival analysis theory, and characterizes the temporal distribution patterns of customer churn. Based on the Markov chain model, it designs a transition probability matrix and prediction framework for customer behavior to capture the dynamic characteristics of state transitions. Using three years of micro-store order data from a certain snack chain brand as a sample, the consumption characteristics of 1,454 high-frequency customers are extracted. Survival analysis is combined to validate the negative correlation between customer system association and survival time, and Sharpley's additional explanatory value is used to analyze the key influencing factors of customer ordering and order cancellation behavior. In the survival experiment, the higher the association degree among units within the customer system, the worse the customer's survival status. Customers with significant fluctuations in consumption amounts exhibit stronger survival resilience, with the longest survival time reaching 23,731 hours. Feature A9 has the most significant positive impact on customer ordering intent, while feature B8 has the most significant negative impact on customer order cancellation intent, making them key variables for precise intervention in corporate marketing decisions.