In recent years, there has been a growing focus on video anomaly detection from unlabeled data, raising the task of Unsupervised Video Abnormal Detection (UVAD). However, most of the existing approaches predominantly focus on global features derived from entire frames, while overlooking local features associated with individual objects. This oversight can result in sub-optimal performance and loss of semantics and explainability of the video scene provided by local features. In this paper, for the task UVAD, we introduce a Global-Local Explainable Network (GLE), which focuses on local features and is based on a Multi-Instance Learning (MIL) method. The proposed approach not only outperforms UVAD state-of-the-art approaches, but also provides explanations about the anomaly by leveraging the rich information within local features. Our experiments demonstrate that GLE achieves the state-of-the-art performance in both detection and explanation. GLE achieves up to 6.22 improvement in the AUC for abnormal event detection on two widely used datasets, UCF-crime and ShanghaiTech. Moreover, GLE offers up to 62.94% enhancement in explanatory capabilities as validated on the X-MAN dataset.