Online shopping has become an indispensable part of people′s daily life in China which is the world′s largest online retail market. In particular, the outbreak of COVID-19 has further stimulated the explosive growth of the “contactless economy” of online shopping. However, with the increasing number of consumers and transactions, online shopping platforms and merchants face challenges, such as inaccurate target consumer orientation, high operating costs in supply chains, and low marketing efficiency. In addition, online consumers are unsatisfied with the speed of delivery and shopping experience. The prediction of online purchase behavior can grasp the tendencies of consumers′ future shopping decisions through the analysis of consumers′ online historical behavior data. It helps merchants better target online consumer groups and optimize marketing strategies. This is an important measure to improve the decision-making efficiency of merchants and the consumer shopping experience. Therefore, the prediction of consumer online purchase behavior has attracted the attention of academic and business communities.Consumers′ purchasing behavior is mainly determined by decision-making psychology, which is easily affected by the complex factors of online shopping scenarios. Therefore, identifying the influence of consumers′ psychological characteristics on online purchase behavior has become an important basis for online purchase behavior analysis and prediction. Most of the existing studies only consider the browsing behavior of users. Few consider consumers′ favorites and carts behavior which can reflect consumers′ decision-making psychology, or explore the influence of consumer involvement level on purchase behavior. Comprehensive analysis of consumer behavior, including browsing, adding to favorites, and adding to the shopping cart, can effectively measure the level of consumer involvement, and describe the psychological process of decision-making. Such analysis improves merchants′ accuracy in the prediction of consumers′ purchase behavior. Besides, existing research mainly focuses on whether consumers will buy certain products, and it does not involve the prediction of shopping quantity and shopping interval of consumers in the future. The prediction of these two types of online purchase behavior, which is much more difficult than the prediction of whether consumers will buy certain products, plays an important role in analyzing consumer value for merchants.Based on consumer intervention theory, this paper analyzes the influence of two types of online consumer behavior, including the shopping quantity and the shopping interval. Consumer intervention refers to the time that consumers spend on searching and processing commodity-related information and the energy that consumers spend on consciously processing commodity-related information and advertisements. It determines the process of consumers′ selection of information categories and making purchasing decisions. In other words, different involvement levels reflect consumers′ decision-making psychology. It will affect consumers′ online purchase behavior and be reflected in their pre-purchase behavior.This paper first defines the dimension of consumer involvement level. It identifies the psychological characteristics of consumers′ online purchase decisions, including the intensity of participation, consumer types, shopping process types, and the number of ad clicks. Then, we explore the influence of consumers′ involvement level on consumer behavior, such as the shopping quantity and shopping interval. A variety of machine learning methods are used to construct a predictive model of consumer online purchase behavior. The comparison among different prediction models shows that the Quantile Regression Neural Network (QRNN) model results in the smallest errors and highest level of fit for predicting online purchase behavior, and the Gradient Boosted Regression (GBR) model results in the best performance for the prediction of shopping interval. Through the reorganization of feature variables, the comparative experiment verifies the effectiveness and rationality of introducing consumer involvement level into the prediction model of purchase behavior. Beyond enriching relevant theories, this study has important practical significance in the accurate prediction of consumers′ online purchase behavior and improvement of service experience.