Smart-Wi-Cache: A Deep Learning Framework for Online Content Caching at the Wireless Edge
Lalfakzuala, Lalhruaizela Chhangte, K. Vanlalawmpuia, Lalengmawia Chhangte · 2025
With the rapid increase in mobile IP traffic, driven primarily by video content, there is a pressing need to improve caching efficiency at the wireless edge to enhance content delivery and alleviate both the backhaul and last-mile network segments. Traditional caching policies, such as Least Recently Used (LRU) and Least Frequently Used (LFU), lack adaptability in dynamic environments and are often insufficient for end-user devices with limited storage and computational power. This paper introduces Smart-Wi-Cache, a deep learning-based online caching frame-work specifically designed for end-user devices at the wireless edge. The core of this approach is LSTM-Replace, an LSTM-based cache replacement policy that predicts content requests at the device level, enabling end-user devices to cache high-demand content. Extensive simulations demonstrate that LSTM-Replace significantly improves cache hit rates compared to traditional caching methods.