Upon the introduction of novel medical therapies, an array of semantically different data is gathered from the participant's cohort. Unsupervised learning is always privileged as a preliminary step for data investigation, to extract valuable information before embarking on the tedious task of data labeling. Clustering emerges as one of the techniques providing a comprehensive overview of exploratory data analysis, facilitating the identification of patient communities. With OptiClust4Rec, we propose a methodology aimed at characterizing the groups, thus providing recommendations for patients undergoing therapeutic treatment. This approach relies on the use of two distinct datasets: the first containing patients' clinical data and the second grouping patients' responses to a questionnaire on their quality of life. Our main objective is to optimize the clustering process and dimensionality reduction, relying on concise metrics and an analysis of the data's topology, in order to effectively label the different clusters and generate relevant association rules.