Abstract
The road from data generation to data use is commonly approached as a data-driven, functional process in which domain expertise is integrated as an afterthought. In this contribution we complement this functional view with an institutional view, that takes data analysis and domain professionalism as complementary (yet fallible) knowledge sources. We developed a framework that identifies and amplifies synergies between data analysts and domain professionals instead of taking one of them (i.e. data analytics) at the centre of the analytical process. The framework combines the often-cited CRISP-DM framework with a knowledge creation framework. The resulting framework is used in a data science project at a Dutch inspectorate that seeks to use data for risk-based inspection. The findings show first support of our framework. They also show that whereas more complex models have a higher predictive power, simpler models are sometimes preferred as they have the potential to create more synergies between inspectors and data analyst. Another issue driven by the integrated framework is about who of the involved actors should own the predictive model: data analysts or inspectors.
Highlights
The road from data generation to data use is commonly approached as a data-driven, functional process in which domain expertise is integrated as an afterthought. In this contribution we complement this functional view with an institutional view, that takes data analysis and domain professionalism as complementary knowledge sources
Differences between inspectors and data analysts It was already assumed that inspectors and data analysts represent distinct knowledge sources, in this case about risks
Current data mining approaches provide hardly any support for incorporating the knowledge of domain experts in the approach to process data explicitly. Often these stepwise approaches assume that knowledge creation and knowledge transfer happens one after the other
Summary
In the past years data intelligence and analytics have sustainably proven their value to many organizations (Hochtl et al, 2016; Janssen and Kuk, 2016; Taylor and Portuva, 2019). Existing data mining frameworks, such as the Cross Industry Standard Process for Data Min ing (CRISP-DM) enable organizations to learn in concert with data (Shafique and Qaiser, 2014; Sharma, Osei-Bryson and Kasper, 2012) These processes propose a structured and iterative sequence of activities, such as problem formulation, data consultation, and analytical model ling. Most frameworks read as a functional chain from data generation towards decision-making They show how data get generated, processed and made ready for those that have to make decisions, either on a po litical or on an operational level (Janssen, van der Voort and Wahyudi, 2017). Professionals may be able to interpret the outcomes, may disagree with them, may misinterpretate them or even neglect them This idea has important consequences for the way data intelligence and analytics can help to improve decision-making.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
