With the increased possibilities to collect and store data ("Big Data"), numerous popular methods have been presented for their analysis. For instance in the medical sector, data analysis is used to provide valuable data for health economic evaluations. However, it was widely overlooked, that most of the newly presented methods are accompanied by non-explainable results, due to so-called "black box" models. This is not acceptable in medical data analytics, including but not limited to the General Data Protection Regulation (GDPR) of the European Union introduced in 2018. Therefore, the importance of explainable analytics in general and in medicine in particular has been discussed and investigated. While one possibility is trying to explain the workings of a black box model, another approach is to introduce advanced analytical methods generating explainable results in the first place. In this thesis we present a new method for explainable analytics to analyze individual patient data in order to generate new findings for an improved future patient care. Furthermore, the new method provides reliable data for health economic evaluations. The new approach is based on an endpoint-oriented clustering approach, developed by Brieden and Gritzmann, forming sufficiently large clusters of patients with similar combinations of their characteristic values. We present a method for the cluster-based analysis of individual patient data to reliably predict the outcome of a patient (e.g. efficacy of a medical intervention). Furthermore, we introduce the newly invented cluster-based survival analysis to predict the "survival" of a patient (e.g. continuance of a treatment). Besides predicting what the outcome of a patient might be, the method provides a unique explanation for the specific prediction, based on individual patient characteristics. Finally, we show the success of the newly introduced explainable method on a real world data set originating from a clinical trial including patients suffering from schizophrenia.    «

With the increased possibilities to collect and store data ("Big Data"), numerous popular methods have been presented for their analysis. For instance in the medical sector, data analysis is used to provide valuable data for health economic evaluations. However, it was widely overlooked, that most of the newly presented methods are accompanied by non-explainable results, due to so-called "black box" models. This is not acceptable in medical data analytics, including but not limited to the Genera...    »