Data mining and knowledge discovery in medical information systems

Hospital Information Management Systems are used in all public and private hospitals. Valuable data which is obtained from these big databases, where updated data are collected continuously during all day, is used for only some necessary reports and queries inside of the corporation, and this data may not be considered for an academic study, because it is not held as a clean data warehouse. In this thesis, it is underlined that this data should be an open source with standard protocols under definite boundaries for academic studies in order to supply the improvement of medical research, and it is mentioned that how this data is processed and valuable patterns are obtained. There are studies about three branches, "medical laboratories", "head and neck cancers" and "nutrition and diets" in this thesis. For these three branches, new information systems have been implemented by using the technologies of MSSQL database and ASP.NET with C# programming languages. Thus, it has supplied to collect clean and processable instances in them and, the technique of knowledge discovery and data mining was able to be applied for these information systems by using different algorithm approaches. In the studies, clustering algorithms from data mining techniques have been implemented with different approaches and firstly, these algorithms have been used to analysis laboratory data sets which necessary permissions have been obtained to use. Secondly, head and neck cancer instances, which necessary permissions have been obtained to use, too, have been analysed by using clustering algorithms with different approaches. Lastly, the data set of nutrition and diets is optimized by hybrid quantum genetic algorithm. Improved clustering algorithms have been compared with self-organizing map, k-means++ algorithms and its different approaches. Moreover, improved hybrid quantum genetic algorithm has been compared with traditional genetic algorithm and quantum genetic algorithm. As a result, the patterns, which have higher accuracy and performance than the traditional approaches, have been obtained.