A Study of Digital Marketing in the Housing Market and the Prediction of Consumer Purchasing Behavior

Abstract

Driven by the wave of digital economy, the real estate industry is facing the transformation of marketing methods. Traditional marketing means have been difficult to meet the precise and diversified consumer demand, and digital marketing means have gradually become an important strategy in the housing market due to its advantages of high efficiency and wide coverage, which has pushed the research on consumer behavior prediction to become an emerging hot spot. This study focuses on the application of digital marketing in the housing market and explores how to predict consumer purchase behavior through multi-model fusion. First, the behavioral data of 13 million users of the real estate e-commerce platform are preprocessed, and the SMOTE and Borderline-SMOTE methods are used to achieve data balance. In terms of modeling, three single models, logistic regression, random forest and XGBoost, are selected for prediction, and the prediction performance is enhanced by Voting fusion algorithm. The empirical analysis results show that the F1 value of Random Forest among the single models is the highest up to 73.68%, while the weighted voting fusion model performs the best in terms of comprehensive performance, with the F1 value elevated to 81.51%, and the AUC value up to 0.6723. The results verify the effectiveness of the fusion model in predicting the consumer purchasing behaviors, and provide data support and technical basis for the implementation of precision marketing in the housing market. The conclusion states that digital marketing tools combined with multi-model fusion prediction technology can significantly improve the conversion rate of home purchase and user identification.