Social media, along with its "event" organization and planning tools, plays an important role in connecting and engaging individuals and groups. These online spaces thrive with multi-faceted 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 talk makes the case for building a data science pipeline to analyze event data and recommend relevant events to 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 next most critical phase, which is feature engineering, storytelling and then on to fuzzy approximate reasoning-based modeling for computing event recommendations.
One particularly desirable property of fuzzy sets is their rich linguistic approximate reasoning ability which allows crunching through numbers and Big Data, while maintaining human interpretability of the built models and predictions. This interpretability is critical in the data science enterprise because data science often require team collaboration and yields results that need to be consumed by people of diverse technical and non-technical backgrounds, who therefore question the meaning of models and emphasize the importance of telling stories from the data. We have evaluated our proposed framework on a real world dataset with more than one million events and 38 thousand users. The proposed approach achieved accuracy of 70% which outperforms other existing event recommendation algorithms.
Mahsa Badami is Ph.D. candidate in computer science at the department of Computer Engineering and Computer Science (CECS), University of Louisville, Louisville, KY, USA. She is currently a member of the Knowledge Discovery and Web mining lab and her research interests include recommender systems, text/data mining and data science, machine learning. She has participated in several recommender systems and data science competitions. In addition, she has published several research papers in various data mining and data science conferences.