Visualizing Survey Data in a Web Interface and Predicting the Reasons Behind Early Leave from Military Service

Brait Õispuu
Estonian Defense Forces regularly carry out surveys among people who are going through military service and collect diverse information about the service as well as the people themselves. The resulting data set has previously already been investigated, but not using automatic methods. This thesis is based on a data set consisting of the answers from people recruited in July and October of 2016. The first purpose of this thesis was to identify which part of the collected information is important in detecting, whether the recruit will be relieved from duty before the standard end date of the service. This was performed using machine learning methods, that have a way of interpreting the importance of a feature used in the created model. Another purpose of this thesis was to create a prototype of a web interface allowing easy and fast overview of potential
problems in different military units as well as track their developments through time.
As a result of this thesis five different machine learning models were designed and their most important features analyzed. Also, the aforementioned prototype of a web interface was created and the utility possibilities were described. In the process of making these models and the web interface, emphasis was put on future compatibility, so that the created tools could be used by the Estonian Defense Forces after additional information has been collected.
Graduation Thesis language
Graduation Thesis type
Master - Computer Science
Mari-Liis Allikivi
Defence year