People routinely convey and receive information via online social media, and those with common interests and traits often form various online social networks. However, the immense amount of complex information on the internet often prevents social network members from efficiently obtaining the information that they seek. This creates the need for a post recommendation mechanism to help social media users find the posts that they want from the all-encompassing content of social media. Popular time periods during which social network members are more likely to browse posts, the relevancy of information to the interests of social network members, and the connections among users often have a direct impact on the acceptance of social network members and posts. If social media managers can effectively identify time periods during which social network members are active, and then interact them during this time, they can maximize post exposure. This study developed a post recommendation and push notification model combining methodologies to gauge post similarity and identify popular post interaction periods. The degree of similarity between posts is determined based on the responses made by social network members, whereas time period of increased interactions are identified based on the numbers of responses made to posts at different time periods. It is hoped that the proposed model can help social network members and social media managers clarify and categorize member preferences and identify critical time periods for social interaction.
Keywords: social network analysis, Recommendation algorithms
Department of Industrial Engineering and Engineering Management, National Tsing Hua University
Before we explain the methodology to gauge post similarity, we first define relevant nomenclature：
Using the procedure above, this study gauged the degree of similarity between two posts based on interactions among social network members. We believe that this approach will facilitate post recommendations to social network members with similar preferences. When a social network member responds to a post, the system will use this approach to recommend similar posts that the social network member has not yet responded to.
Before we explain the methodology to identify time period of increased interactions, we first define relevant nomenclature:
This methodology uses the procedure above to identify the time periods of increased interaction. Posting information updates during these time periods will maximize exposure among users of a social network. The information that results from this methodology can also be helpful when gauging post similarity so that the social network automatically recommends posts and gives new information to social network users when they are more active.
Today, social network members browse social media websites every day to receive the latest news and read articles that they are interested in. However, due to a glut of information, social network members cannot be sure whether the post they are beginning to read will meet their expectations in terms of content and quality, let alone whether they should read more of the posts. As a result, they may waste a lot of time reading posts that are not helpful to them and even have difficulty deciding what they are interested in. To resolve this issue, we developed methodologies to gauge post similarity and identify time periods of increased interaction. This knowledge can then be applied to post recommendation functions. This will enable social media systems to automatically recommend posts that will be meaningful to users at opportune moments. The contributions of this study are as follows:
- Methodology to gauge article similarity: The types of responses made by social network members to various posts are first collated to derive response distributions that can be used to calculate the degree of correlation between posts. This provides reference for post recommendations.
- Methodology to identify time periods of increased interaction: The total response frequency for all posts at all time periods is obtained. Then, the ratios of the response frequencies are derived to obtain the time periods of highest interaction.
1. Perny, P. and J. D. Zucker. Preference-based Search and Machine Learning for Collaborative Filtering: the “Film-Conseil” recommender system. Information, Interaction , Intelligence, 1(1):9-48, 2001.
2. Herlocker , Jonathan L., Joseph A. Konstan , Loren G. Terveen , John T. Riedl, (2004) Evaluating collaborative filtering recommender systems, ACM Transactions on Information Systems (TOIS), v.22 n.1, p.5-53, January 2004.
3. Golbeck, Jennifer (2005) “Computing and Applying Trust in Web-Based Social Networks,” Ph.D. Dissertation, University of Maryland, College Park.
Yuan-chen Liao, Annette Li-an Chiu＊, Yu-jui Wan ,Wei-ting Liao, Adam Hou (Department of Industrial Engineering and Engineering Management, National Tsing Hua University Graduation project:”Social Network-Based Recommend System” Award for excellent performance )recommend_system