Cloud Computing-based Parallel Recommendation Algorithms for Social Networks

Abstract

The rapid development of Internet technology has brought convenience to various industries
and also produced a large amount of information redundancy. Internet users are facing severe
information overload problems. The search engine for information retrieval has alleviated the
information overload to a certain extent, but with the surge in data volume and the
development of recommendation systems, major social networking sites, operators,
e-commerce, etc. have launched their own personalized recommendations Products, which
brings the gospel to those “difficult to choose.” Nowadays people are pursuing more
personalized and faster high-quality services. Therefore, the implementation of a
recommendation strategy that can be applied in the state of big data has become an important
means to solve information overload and improve service quality. The birth of cloud
computing platform has promoted the development of recommendation systems. Traditional
social network parallel algorithms have shortcomings in big data processing. When using
check-in records, they cannot make full use of the preferences, locations, and social network
information implied by check-in information, resulting in low accuracy. Based on this, a cloud
computing-based parallel recommendation algorithm for social networks is proposed, and the
Debogone algorithm is designed. Debogone algorithm design is realized by feature extraction
algorithm. Based on the user’s historical preferences, the user’s social relationship is
comprehensively considered, and the user’s range of activity is used as the constraint point to
implement the user’s interest point recommendation. Through experimental comparison, it is
proved that Debogone algorithm has high design accuracy, high stability, and has the
significance of popularization. This paper studies a parallel recommendation algorithm for
social networks based on cloud computing. In this paper, the Epinions public data set is used
as experimental data and experimental objects. Through data analysis, the prediction accuracy
rate is 85%, and the recall rate is 59%. Compared with the traditional model, the prediction
effect is improved. Model-3’s logistic regression parallel voting model.The model’s algorithm
accuracy rate is 83.4% and the recall rate reaches 62.3%. The accuracy of discrimination is