Content-Based Recommender for U-13 Platforms Using Weighted Keys Vector Transformation for Age Classification and Topic Modeling

Abstract

This paper details a recommendation system for social media posts directed toward parents and children under the age of 13. The content-based recommender uses two main features to draw similarities between posts: predicted age and predicted topics. We detail a vector transformation algorithm and classification techniques that provide 81.7% accuracy predicting age ranges under 13 and 90.2% accuracy assigning content topics to posts. These lassification methods allow us to make better content suggestions on U13 platforms.