In this article we have presented a classification model for metadata extracted from tweets using collaborative filters, proposing a metadata recommendation system at the knowledge level. Due to the limited resources available for tagging in social media, the quality of metadata used for this practice often poses interpretation issues when selecting, resulting in a loss of engagement due to flaws in term formation or a lack of understanding of the shared metadata within the social network. In this study, we proposed a selection algorithm that filters and classifies metadata based on metrics that measure the collective intelligence aggregated in the metadata. We then index knowledge indexes that are readily identifiable visually. The proposed model involves two main steps for classifying and recommending metadata.