In the context of the rapid development of the Internet of Things (IoT), although a large number of sensing devices have been deployed and vast amounts of data are available, single-view data often suffers from limited perspectives and high uncertainty, leading to issues such as incomplete information and susceptibility to noise, which undermine the reliability of model decisions. To address these challenges, this paper focuses on the problem of distributed trustworthy decision-making using multi-view sensing data, and proposes a distributed and reliable classification method based on multi-view fusion. The proposed method incorporates evidence-based uncertainty modeling and fully accounts for both the consistency and complementarity of multi-view information at the decision-making level. Specifically, in the single-view learning stage, deep neural networks are employed to extract evidence from each view. Based on subjective logic, Dirichlet distributions are used to construct credibility models for each perspective within the Dempster-Shafer evidence theory framework. In the multi-view fusion stage, an improved Dempster combination rule is used to effectively integrate the credibility models from different views. Additionally, a conflict-aware inconsistency measure is designed to assess decision-level disagreements across views. Experimental results on six datasets demonstrate that the proposed method outperforms existing mainstream feature-level and decision-level fusion algorithms in both classification accuracy and uncertainty modeling, under both conventional and challenging scenarios, thereby validating its effectiveness and advantages.