Article Preview
TopIntroduction
The number and types of services continue to increase due to the development of technologies like cloud services, internet of things (IoT) services, mobile services, and microservices. Service recommendation technology emerged to identify a service that meet individual users’ needs (Qi, 2023; Zhang, Y., Yin, C., et al., 2021). Service recommendation is generally used to model a user’s interest by analyzing their historical behavior. Thus, a service can be recommend that meets a user’s preference requirements (Zhang, Y., Cui, G., 2021). In a service recommendation system, the quality of service (QoS) should be predicted when identifying a service that meets a user’s quality requirements. Therefore, accurate prediction of the QoS value (Zhang, Y., Wang, K., He et al., 2021; Qi et al., 2020; Qi et al., 2022) prior to the user call is an important step in service recommendation (Zhang, Y., Zhang, Yan, 2023; Zhang, Hu, Zhang, 2021; Cui et al., 2020).
Edge computing (Zhang, Pan et al., 2021; Zhang, Cui, Zhao et al., 2016; Zhang, Zhao, Deng, 2018) has introduced new problems to existing recommendation technology. When an emerging technology attracts new users, a recommendation system is often used to provide personalized recommendations for new users. The user cold start problem is monumental in recommendation systems (Wang, 2021; Seth & Mehrota, 2021; Amamou et al., 2016). Existing methods to address the cold start of users can be divided into three types. The first, statistical methods, includes the mode method and average method in statistics. These can alleviate the user cold start problem to a certain extent. This type of method is easy to implement; however, it has poor prediction accuracy. The second, demographic methods (Rashid et al., 2008; Liu et al., 2018), uses new users’ completed questionnaires to profile users. This type of recommendation method achieves better prediction accuracy than statistical methods; however, it involves user privacy and may lead to a decline in user favorability and user viscosity. The third includes methods based on foreign information sources (Wang et al., 2017; Sahebi & Brusilovsky, 2013). These have good prediction accuracy and recommendation effects. Still, the foreign information sources infringe more on user privacy.
This article focuses on the cold start problem in web services and edge services, including at the protection of user privacy. A recommendation algorithm is proposed based on matrix factorization to solve the cold start problem of users. There is a certain correlation between user preferences and cultural background; therefore, this study introduces Hofstede’s cultural dimensions theory (Hofstede et al., 2010). First, it establishes the connection between two users. Second, it ensures QoS for cold start users’ stability of prediction accuracy. The experimental results show that the proposed method largely solves the problem of predicting the accuracy of cold starts by users. The main contributions of this article are presented as follows: