In recent years, advancements in vehicular communication and networking have played a pivotal role in accelerating the development and adaptation of autonomous vehicles. Within this framework, Information-Centric Networking (ICN) has emerged as a preferred architectural choice. The intrinsic caching capabilities of ICN enhance content delivery efficiency, making it suitable especially for devices with limited resources. However, despite caching being a crucial trait in ICN-based networks, there remains a difficulty in efficiently reducing content retrieval time when only the most essential contents can be stored due to limited cache capacities. In this paper, we propose a novel two-level cluster-based caching approach that considers contextual information of the users and Cache Benefit Score (CBS) for placing the content. The proposed scheme uses Neural collaborative filtering (NCF) in the Federated learning (FL) framework to predict the content popularity score which is then used to calculate the CBS for content placement. The effectiveness of the proposed scheme has been evaluated in the Icarus- an ICN simulator, with some recent caching strategies. Results show that the proposed caching scheme notably enhances efficiency, reflected in cache hit ratio, content retrieval delay, and average hop count.