Abstract: Personalized product recommendations based on user profiles is a widely applied method of precision marketing, relying on computational power, model algorithms, and a multi-channel data foundation. It is favored by e-commerce platforms for its automation and precision. However, as brands increasingly influence consumer purchasing decisions, the lack of brand characteristics in the personalized recommendation process hinders further improvement of recommendation effectiveness. On the other hand, unlike the tangible attributes of products, brands largely exist in the minds of consumers, making brand features not as visually apparent as product features. Therefore, this paper is dedicated to extracting brand features from the user's perspective and then building a brand preference identification model that meets consumers' personalized needs. Initially, the paper utilizes the BERT neural network model to generate brand feature vectors through brand comment corpora on Weibo and calculate perceived brand similarity. Subsequently, by conducting experiments to acquire consumer brand preference data, a user association matrix for brands is generated based on overlapping consumer brand preferences. Finally, through machine learning algorithms, different consumer tendencies in terms of brand perception similarity and user association are identified, leading to the construction of a mixed collaborative model combining both dimensions, ultimately achieving personalized brand preference recognition for consumers. The research finds that consumer perceived similarity and user association can reflect consumer personalized brand preferences to a certain extent, with different machine learning algorithms having various advantages in identifying brand preferences. This paper provides significant theoretical and practical guidance for personalized recommendation based on brands and innovation in intelligent marketing.

Key words: brand preference, neural network, social network, user reviews, perceived similarity, user relevance