This document describes the development of a data-driven learning interest recommendation system on MOOCs. It discusses using students' activity logs on MOOCs, such as video clickstreams, to analyze their interests and provide recommendations to help students. The system, called Videomark, analyzes video activity data to identify keywords and popular video segments for each week's content. It displays these as a keyword cloud and allows students to click keywords to access related video sections. An experiment on a networking security course found Videomark improved students' learning performance, engagement and experience compared to not having the recommendation system. Future work could include generating keywords in other languages and improving the user interface.