Prediktiv analys i människans tjänst

• Main Authors: Elfving, Markus, Althin, Tom
• Format: Others
• Language: Swedish
• Published: Uppsala universitet, Avdelningen för visuell information och interaktion 2019
• Subjects: Predictive Analysis; business intelligence; Contextual Design; visulization; Prediktiv analys; beslutsstöd; visualisering; Other Engineering and Technologies not elsewhere specified; Övrig annan teknik; Human Computer Interaction; Människa-datorinteraktion (interaktionsdesign)
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-388603

Predictive Analysis is a process for extracting information from large amounts of data and using it to make qualified predictions about future results. While previously the lack of available data has been a challenge within the field, big questions today are instead how to use the results, and the way in which these are presented in order for the user to be able to take advantage of the information. The purpose of this thesis has been to create hypotheses for how predictive analysis can be used in practical decision-making contexts, whereby the decision- maker is under time pressure, especially with regard to how the result can be visualized. This has been done through a case study at the Uppsala Ambulance Monitoring Center. The method used for the study is called Contextual Design, which has helped create an understanding of the users and the system they work in. Using this understanding, a prototype has been created, which has been tested on the users to see how well they have been able to interpret the information that has been visualized. Predictive analysis has proved to be helpful primarily in less urgent cases and to help the decision maker to differentiate matters similar to each other. For visualization of the predictive results, it has been found that these is better shown as a comparison between the user's decision hypothesis and historical decision results rather than only as an absolute value. Furthermore, it has been found that a high degree of transparency in the information on which the results are based is preferable, but that it is important that clear explanations are given for the results shown.