The recommender systems are increasingly being more inrested in many fields of the life, it is increasingly effective in finding information to suggest to users in big data systems. The multi-criteria recommender systems are always researched and improved to suit the diverse requirements of data and user preferences today. The calculation to make a reasonable decision is required for the multi-criteria consulting system. Many operations have been applied to decision making. Most traditional recommender systems often use average operations to calculate useful values used in decision making. In this paper, we offer a new approach to develop decision making based on the level between criteria, sets of criteria. In an information system, data always have an interactive relationship that represents the intrinsic values in the system. If we do not fully calculate these values for decision making, then the decision making will not be fully effective. We build a multi-criteria recommender model with both Item-based and User-based (compare and evaluate the results with some existing models, applied decision-making operations) with different levels. interactions between criteria and different sets of criteria for decision making. Experiments show that when using different interaction values, the results are different and the decision-making efficiency increases with the level of interaction, contributing to improving the quality of the current multi-criteria recommender system.