In recent years, the rapid development of e-commerce live streaming has made the impact of user perception experience on purchasing behavior become a hot topic of research. This paper explores the relationship between user interaction frequency, purchase intention, influencer impact, and purchasing behavior using a Support Vector Machine (SVM) model. By constructing multi-dimensional feature combinations and conducting experimental analysis, the results demonstrate a significant correlation between user perception experience and purchasing behavior. Specifically, when combining interaction frequency, purchase intention, and influencer impact, the SVM model achieved a classification accuracy of 90%. Compared to traditional classification algorithms such as decision trees and random forests, SVM exhibited higher accuracy and F1 scores when handling complex nonlinear data, confirming its superiority in predicting user behavior in e-commerce live streaming scenarios. This study not only reveals the behavioral patterns of users in e-commerce live streaming but also highlights the potential of machine learning algorithms in user behavior analysis. The SVM model effectively captures the key features of user behavior and provides theoretical support for optimizing personalized recommendation systems and improving purchase conversion rates for e-commerce platforms. The paper further discusses the advantages of SVM and its practical application in e-commerce contexts, while offering directions for future research.