A Cognitive Prioritization for Reports Generated in Resource Constrained Applications
Abstract
In many resource constrained web/cloud applications, users are given the ability to generate different kinds of reports after selecting some criteria for each report to be produced. Such criteria could, for example, be filters on certain report attributes. With limited computing resources, it is critical to prioritize the reports requested by users at any certain period of time. Then, that prioritization can be fed into any of the known web services scheduling algorithms. In this paper, we provide a novel cognitive prioritization approach that takes into consideration the free-form user-input text about the criticality of the reports as well as their aforementioned structured attributes/filters. Our method consists of a predictive model that uses the structured data to predict the report completion time as well as another text-mining model that uses the user's text to weight its importance. Then, both outputs are combined with the user profile to come up with the final prioritization. We apply our methodology to real data taken from the report generation of a real-world application of IT service deal pricing and show results that illustrate the effectiveness of our approach.