Using feature selection or sample selection?

I'm Implementing an article in Matlab( the article is attached). You can get the dataset in this link: https://sites.google.com/site/tmisdata/ The article's framework includes: 1-Building the social Network for stakeholders, 2-Clustering the graph, 3-Finding the most probable ideas for each cluster. In phase 3 my question is whether to choose feature selection or sample selection for the dataset? I'll be thankful for your point of view.