Article Preview
TopIntroduction
Cross-border e-commerce (CBEC) is one of the crucial ways or platforms for modern consumers to shop (Zhu et al., 2019; Kawa & Zdrenka, 2016). Relying on the progress of blockchain, CBEC has established a substantial industrial chain and innovative sales methods, making it convenient for people in different countries or regions to get more goods. However, different consumers have different needs and purchasing goals, and mastering consumers’ purchasing preferences as soon as possible can profoundly impact the strategic decisions and operations of CBEC.
The consumer preference model is a mathematical or statistical model that describes and predicts consumers’ purchasing behaviors and choices (Chung & Rao, 2012). It can help e-commerce businesses better understand consumers’ needs and preferences, optimizing product positioning, marketing, and service strategies. The consumer preference model also has its characteristics (Zhu & Li, 2018). First, consumer preferences are diverse, and each has unique preferences and needs. Considering consumers’ differences, a comprehensive and accurate preference model may need to cover many factors and variables. Second, consumer preferences may change over time as the environment and other factors change. Therefore, the established model must be adaptable and able to capture the changes and trends in consumer behavior at any time (Al-Alawi & Bradley, 2013). Then, the construction of the consumer preference model requires too much data support, and these data may involve multiple dimensions, such as personal information, purchase history, and social media behavior. The collection, collation, and analysis of data is a complex process. In addition, various factors often affect consumers’ decisions, and sometimes, consumers’ decisions may be uncertain, leading to a certain degree of uncertainty in the predicted results of the model (Xie et al., 2021; Viciunaite & Alfnes, 2020).
According to the above characteristics, there are many difficulties in constructing the consumer preference model. First, the consumer preference model data cost is high, and many complex and valuable features must be extracted. Steps such as data preprocessing and feature engineering require careful processing to guarantee accuracy and reliability (Nguyen et al., 2019). When building a consumer preference model, appropriate mathematical or statistical models need to be selected to describe consumer behavior and choices. Second, consumers’ decision-making may be influenced by multiple factors, such as personal characteristics, environment, and competitors. Finally, because many factors influence consumer behavior, sometimes consumers’ decisions may be irrational (Yenipazarli, 2019). Therefore, the model’s predictive accuracy may be limited to some extent. The consumer preference model has the characteristics of diversity and dynamics, while data acquisition, processing, model selection, and prediction accuracy are the main difficulties in constructing the consumer preference model. It is necessary to apply data science comprehensively, statistics, and industry experience to overcome these difficulties and better meet the market demand and the individual demand of consumers (Ma et al., 2019).