Fantahun Mesganaw (Professor)Abdela Emamu2018-11-272023-11-292018-11-272023-11-292011-06http://etd.aau.edu.et/handle/123456789/14529Today’s world is encountered a lot of social, economic and political challenges. Among these health challenges take major part, three health related Millennium Development Goals (MDGs) were set and have been implemented to tackle these problems. There are various health problems. Emergency medical health problem is one of them which is critical and affect many people in the world, especially in developing countries. As statistics shows, in Africa region in general and Ethiopia in particular, the emergency medical case situation is still worse and needs special attention particularly Road Traffic Accident (RTA). A lot of demographic and clinical (related) data is recorded about patients who come and receive treatment in the emergency medical service unit of the hospital. As these data are getting larger and larger, there can be a probability in which hidden, implicit and non trivial knowledge exist within these data. So far it is recognized among scientific scholars traditional statistical methods might not be good enough to discover such hidden knowledge from large and complex volume of data. This is where data mining becomes very important to mine such hidden, complex, necessary data to generate vital knowledge. There were various activities carried out throughout the research work based on CRISP (Cross- Industry Standard Process) data mining methodology. The source data for this research purpose was collected from Tikur Anbessa Specialized Hospital Emergency Medicine Registration Database which has 5708 instances. Important patterns and variables related to cause of visit were identified. Data preprocessing activities were made which include handling missing, inconsistent and noisy values that took most of the research time. Appropriate data mining techniques or functionalities were selected and employed for the research work. These are classification and association rule mining. For classification purpose decision tree classification with J48 algorithm and rule induction with PART algorithm were employed. On the other hand, Apriori algorithm was used for association rule mining purpose. Weka(Waikato Environment for Knowledge Analysis) 3.6.0 version data mining tool was used for model building and experimentations.Four experiments for decision tree classification models, one experimental model for rule induction, and ten experiments for association rule mining were done by varying parameters. Among PART and decision tree classification models experimentation made, pruned decision tree with default confidence factor (0.25) has slightly better performance, accuracy measures and generated rules than the other four models. On the other hand, many association rules with acceptable patterns by domain experts were obtained. Interesting patterns were generated by using classification and association rule discovery models related to the problem domain. Over all the researcher tried to use techniques in discovering patients cause of visit in the emergency unit.enMining Emergency Medical DataThesis