I’d like to report on our demonstration paper at WWW 2017, focusing on Spark-based Big Data Analysis of Semantic IFML Models and Web Logs for Enhanced User Behavior Analytics.
The motivation of the work is that no approaches exist for merging web log analysis and statistics with information about the Web application structure, content and semantics. Indeed, basic Web analytics tools are widespread and provide statistics about Web site navigation at the syntactic level only: they analyze the user interaction at page level in terms of page views, entry and landing page, page views per visit, and so on. Unfortunately, those tools do not provide precise statistics neither about the content and semantics of the visited pages, nor about the actual reactions of the users to the actual content (instances) he is shown.
With our work we demonstrate the advantages of combining Web application models with runtime navigation logs, at the purpose of deepening the understanding of users behaviour.
We propose a model-driven approach that combines user interaction modeling (based on the IFML standard), full code generation of the designed application, user tracking at runtime through logging of runtime component execution and user activities, integration with page content details, generation of integrated schema-less data streams, and application of large-scale analytics and visualization tools for big data, by applying data visualization techniques that build direct representation of statistics on the IFML visual models of the Web application.
The paper describing the approach is available in the WWW 2017 proceedings.
The video of the demo is available on YouTube: