Network Science has become a flourishing interest in the last decades with the Big Data explosion. To improve multidimensional data Recommendation Systems, multiplex graph structures are useful to capture various types of user interactions. We propose a graph database approach to compute multiplex graphs which helps both manipulating dimensions in a flexible way and enhancing expressiveness with algebra to express manipulations on the multiplex graph. Applied operations rely on graph algorithms to predict user interactions. We compare our approach with the traditional matrix approach with Random Walk with Restart. The study shows that combination of scores from layers of multiplex graphs provide important insights into user preferences, with most configurations outperforming traditional matrix methods. This approach provides a comprehensive analysis of multidimensional recommendation strategies in multiplex graphs, which provides capabilities of managing different dimensions for queries, paving the way for more sophisticated and customized recommendation systems and its explicability.