<p>In the current analysis of uncertainty in decision-making, multi-attribute three-way decision-making (TWD) is an important theme and an effective decision-making tool. However, with the development of the era of big data and information, single source data is difficult to better handle uncertainty problems. Moreover, compared to single valued data, interval valued data is more effective in describing the uncertainty and volatility of data. Therefore, it is necessary to study multi-source interval value information systems (MsIVIS). This article proposes a three-way decision model for multi-source interval value fuzzy decision information systems (Ms-IVFDIS). Firstly, we extract the initial matrix of attributes from the obtained multi-source interval value information system, further converts the initial matrix into a support matrix and a Basic Probability Allocation (BPA) matrix, and obtains the final single source interval value information system through hierarchical Dempster-Shafer (D-S) evidence theory fusion. Subsequently, a new interval value based TWD model is proposed, which defines a new similarity class of interval values and constructs corresponding conditional probabilities. The relevant regions are divided into three parts: positive, negative, and boundary. The relative utility function reflects the preferences of decision-makers. Finally, the effectiveness, rationality, and stability of the model were verified through nine UCI datasets and case analysis.</p>