This paper explores the effect of forecasting methods on the Optimistic Follow The Regularized Leader (OFTRL) caching policy. It has been theoretically proven that the performance of OFTRL improves with accurate forecasters. However, the forecasters were portrayed as black boxes. In this paper, we do not treat forecasters as a black box but rather use recommender systems and temporal convolutional networks (TCN) implementations as forecasters for OFTRL in the caching context. The said forecasters predict the next file requests directly from the request trace rather than monitoring popularity. Using request traces extracted from the well-known MovieLens dataset, it was found that incorporating a forecaster into an optimistic learning algorithm is not straightforward. In fact, it was found that simpler approaches such as one that predicts the most requested file, can outperform more complicated deep learning models in terms of regret even if they possess a lower accuracy.

The advent of wireless networks such as content distribution networks and edge computing networks calls for more effective online caching policies. Traditional policies lose performance since these new networks deal with highly non-stationary requests and frequent popularity shifts. Consequently, a new framework called Online Convex Optimization (OCO), which does not assume the request pattern, has recently been used to tackle the online caching problem. Besides, in many practical scenarios, a request prediction of unknown quality is available. This paper will leverage that and proposes a new online caching policy that uses these predictions. This policy will use the Optimistic Online Mirror Descent (OOMD) algorithm to solve the OCO problem. The policy will still obtain the same regret bound as its non-optimistic counterpart up to some constant even if the predictions are not accurate. The performance of the proposed policy is evaluated and compared with previous OCO-based policies with the use of trace-driven numerical tests.