The Personalized systems are generally based on collecting and exploiting users’ preferences by exploring their traces’ data. Actually, they find users who have similar attributes, cluster them, and then applying algorithms using the subnets. The similarity between users compares their profiles including their attributes. The adding of tags to enrich the user profile must take into consideration the long and short term criteria of the user’s attributes that change over time. In this paper, we present a tag-based profile enrichment approach by adding a time score describing the short and long term criteria of the attribute. Then we use graph analytics to draw clusters of users by inspecting similar tags. Our approach helps companies to make their predictions and conclusions. The datasets of patients’ images ChestX-Ray14 have been conducted to evaluate the effectiveness of our approach.

Summary To support health professionals in making decisions, CDSS are developed to manage the patients' EHR, improve the way of diagnosis, and treatment of diseases. The process of analyzing EHRs is based on reading free‐text notes. However, it spends time and physicians' efforts. In this case, the most used solution is describing the EHRs with shortcut tags, representing pathologies or diseases, which are well‐defined and meaningful information. Still, this solution remains insufficient. The exploration of the relationship between those tags, the EHRs and their belonging patients will improve the analysis and then the CDSS. In this paper, we present a graph‐based tag recommendation approach that suggests relevant tags (diseases and pathologies) by analyzing the tagged medical images. We use graph analytics to generate graphs of tags, patients, and images by inspecting similar medical images descriptive. We have also created sub‐communities of patients with the same diseases by applying the Louvain clustering method. The tag recommendation aims to enhance the computer‐aided diagnosis in medical imaging. The tag recommendation approach will allow radiologists to detect and interpret invisible diseases of the underlying anatomical structure. It will also help in early revealing and diagnosis. The dataset ChestX‐Ray14 has been conducted to evaluate and test the accuracy and effectiveness of the proposed approach. Future perspective will focus on the deployment of our proposal within a Moroccan e‐health project.