Explainable Machine Learning and Mining of Influential Patterns from Sparse Web

Carson K. Leung, Fan Jiang, Yibin Zhang

Abstract: With advancements in modern technology in the current era, very large volumes of big data have been generated and collected in numerous real-life applications. These have formed a connected world comprising webs of agents, data, people, things and trust. Some of these webs have also emerged in health and smart living. As valuable information and knowledge is embedded in these rich sets of webs, web intelligence is in demand. In this paper, we focus on a data science task of web usage mining. In particular, we present a web intelligent solution to conduct explainable machine learning and mining of influential patterns from sparse web. It provides a compressed representation of sparse web, discovers influential websites and/or web pages that are frequently browsed or surfed by web surfers, and recommends these influential websites and/or web pages to other web surfers. Evaluation results show the effectiveness (especially, in data compression), interpretability and practicality of our solution.