| dc.description.abstract | Knowledge Discovery (KD) and Data Mining are two well-known and still growing fields that, with the advancements of data collection and storage technologies, emerged and expanded with great strength by the many possibilities and benefits that exploring and analyzing data can bring. However, it is a task that requires great domain expertise to really achieve its full potential. Additionally, it is also an activity which is done nowadays mainly by data analysts and scientists, which most of the times knows little about specific domain subjects, like in the healthcare segment, for example.
The term Applied Data Science (ADS), recently introduced, focus on creating means, by using analytical methods and applications, for facilitating the daily life of domain experts. Thus, in this research, following an ADS orientation, we propose means for allowing domain experts from the healthcare segment (e.g. doctors and nurses) working in the Wilhelmina Kinderziekenhuis (WKZ), to also be actively part of the Knowledge Discovery process, focusing in the Data Preparation phase, and use the specific domain knowledge that they have in order to start unveiling useful information out of the data. Hence, a guideline based on the CRISP-DM framework, in the format of methods fragments is introduced to guide these professionals through the KD process, focusing in the data pre-processing stage. In order to build the model, an extensive literature review was performed, followed by interviews which aimed to understand what domain experts actually knew about KD, and what should be feasible for them to do when addressing an analytical problem. In addition to that, also to understand what types of problems domain experts would be dealing with in their daily routine, a data quality assessment from the available information within the databases from the WKZ was performed.
Regarding the evaluation of the proposed solution, five meetings with domain experts were held, where the model has introduced and extensively explained, and two case studies representing a real analytical project (using real data) were performed. The findings of this study were acquired by means of a survey, which extracted their opinions about the interpretability (understandability and accuracy), ease of use, perceived usefulness, and intention to use the MAM. The results (described in the previous section) showed that domain experts were very much satisfied about both understandability and accuracy of the model, as well as with its perceived usefulness. Additionally, regarding the model’s ease of use, that is the effort it took to both understand it and to follow it, although not optimal the ratings were above average, which is considered to be normal since they were seeing and experiencing it for the first time. Finally, most participants said that they have the intention to use the model in future activities. | |
