Developing of a Learning-based System to Assist Treatment Process of Arrhythmia Patients

Abstract

The transformation process of the data kept in data warehouse into usable information for decision support system is very important. In this process, it is necessary to reveal the useful data that can meet the needs of the developed system. For this purpose, the study of data mining should be done on data warehouse. In this article, the phases of developing and applying a learning-based system in order to be used in the treatment process of arrhythmia patients were presented. Firstly, the data collected in the data warehouse was transformed into information through using data mining methods. In the data mining studies, the nearest neighbour algorithm (kNN) which is one of the machine learning algorithms was used. The process of selecting training data among the data stored in database was realized by a specialist through a web-based practice. In the process of composing training data, the evaluation results of the system about learning level were shown to the specialist in an online way. Thus, the generalizing power of the model made up by a classifier on training data was measured. Besides, to determine and filter the noisy data in the data warehouse, the quality of the signal taken from the patient and the evaluations of the specialist were used. In this way, the classifier used was contributed to form a suitable hypothesis on training data. By finding the proper hypothesis, the critical situations of the patient are conveyed to the doctor of the patient by the improved system within seconds. Thus, the patient is provided to be constantly kept under supervision free from location.