Towards a Flexible Experience of Data Provenance Summarization

Abstract

In complex data analysis or applications, it is often necessary to collect, aggregate and manipulate large amount of data from multiple sources. Data provenance and their approximated summarizations have been proven to be helpful for recording and understanding user behaviors in these aspects. The growing urgency for providing timely feedbacks and the increasing need to explore more possible summarization options create a much bigger new challenge: exploring more extensive state-space under shorter response time constraints and fulfilling more complex requirements at the same time. In this work, we propose that this can be achieved by relaxing the greedy strategy adopted by existing approaches and by introducing a more flexible optimization strategy based on incremental and adaptive sampling. Our evaluations show that, compared to existing approaches, summarization processes guided by our strategy may produce more flexible and satisfying service data provenance summarization results at smaller temporal and spatial resource costs.