A case study for intelligent event recommendation

Social networks, along with their “event” organization, planning, and sharing tools, play an important role in connecting and engaging individuals and groups. These online spaces thrive with multifaceted activities and interests which give rise to rich content and user interaction that often crossover to the world of events. For these reasons, the data trails associated with “events” in the virtual world can be complex and challenging to understand and predict. This paper presents our efforts to build an interpretable framework to analyze event data and recommend relevant events to social media users with different preferences. The datasets for this challenge were provided by a competition on Kaggle. We conduct an extensive data analysis and exploration to help gain a better understanding of the data. We then proceed to the critical phase of feature engineering, storytelling and modeling for computing event recommendations. We explore fuzzy approximate reasoning for modeling because of its rich linguistic expression ability which allows handling uncertainty, while maintaining human interpretability of the built models and predictions. This interpretability is critical in the data mining enterprise because data mining often requires team collaboration and yields results that need to be consumed by people of diverse technical and non-technical background. Such teams tend to question the meaning of models and emphasize the importance of telling stories from the data. We evaluate our event recommendation system on a real-world dataset with more than one million events and 38,000 users. The proposed methodology achieved 70% accuracy, outperforming existing event recommendation algorithms.