Eureka! Institutional Repository

One of the biggest problems in the health sector is patients not showing up for scheduled appointments. This phenomenon has a significant impact, both on patients due to the low quality of service provided by health structures, and on the functioning of the structures at an organizational, administrative, financial level. To deal with the problem, one way is to predict the patients who will not show up for the scheduled appointments. In this way, appropriate personalized measures will be taken for each patient, such as a reminder of the appointment. This can be achieved by using knowledge mining methods on medical appointment data. In this thesis, the purpose is to compare knowledge mining methods and specifically to compare classification algorithms through an experimental process. To perform the experiments, k-NN, Naïve Bayes and decision trees (C4.5) algorithms were used, which were applied on 2 freely available datasets with medical appointments. In addition, the possibility of improving the classification results by applying the SMOTE oversampling technique to the above datasets was investigated. WEKA free knowledge mining software was used to pre-process the data and run the experiments. After the experimental process, the results were compared and conclusions were drawn regarding the performance of the classification algorithms, as well as the degree of influence of the SMOTE technique on their performance.