A recommender system provides personalized recommendations to theuser in a large space of possible options. Most of the existing recommender systemsutilize users profile data for recommending an item. However, these conventionalrecommender systems are under scrutiny due to strict personal data protection lawsaround the world. Additionally, we investigated the performance of a graph-basedapproach for the recommendation systems. In this thesis, we present a graph-basedscalable and novel approach for the recommendation which doesnt depend on the profiledata for predictions. We evaluate our method on an extensive transaction datasetfrom the retail domain (700k transactions with 150k different items) and compare itto a baseline. The proposed approach relies on knowledge graph embeddings. Duringour evaluation, we have used two knowledge graph embedding algorithms. Thesuggested method first applies Pyke, a knowledge graph embedding algorithm with aclose to linear runtime complexity. Later Pyke was replaced by a convolutional complexknowledge graph embedding algorithm, Conex. The evaluation results suggestthat Conex fits better than Pyke to the approach. Our implementation is open-source,and it is available at https://github.com/nil9/Master-Thesis.