(VRU) is critical to enhancing the interaction process to detect, categorize, and predict patterns and behaviors to decrease fatal accidents for VRUs. Multiple challenges are unsolved to address efficiencies in understanding potential risky behaviors from VRUs and connected vehicles. Ample possibilities to expand research exist in different domains, under the data perspective, modeling behaviors, increasing accuracy during all the interactions, and considering other exogenous variables, such as crime rate or environmental conditions. Despite technological advances, recent literature highlights rising traffic casualties related to vulnerable road users (VRU), such as pedestrians, cyclists, motorcycles, and animals. Therefore, designing a robust, resilient, and reliable system for end-to-end VRU protection is critical. Unsupervised and Deep Learning were combined as techniques to predict VRU injury risk levels. A proposed framework and the “V2X-VRU Dashboard” App Visualization are developed. Our project considered data gathered from multiple sources in real-time. There were two primary sources: 1) New York City DOT and 2) the NYC Open Data for Motor Vehicle Collisions – Person gathered during 2021. Our study shows a combined methodology of neural network-based PCA to ingest real-time data (offline), process, and classify the risk levels of fatalities. It is an empirical and straightforward method to define the principal components of neural network-based PCA.