This paper addresses the single-agent continual area sweeping problem by providing an analysis of prior works and proposes a data-driven approach for more realistic scenarios of event occurrences. The single-agent continual area sweeping problem is a non-uniform extension of the single-agent area coverage problem where there are areas on a map with events occurring with Bernoulli probability per time step. In particular, thedistribution of event occurrences on the map is unknown to the agent. The goal of the agent is to maximize the number of events detected as well as minimize the average event detection time. This paper focuses on the variant of the continual area sweeping problem where the number of events per region is bounded and expands on prior work by removing assumptions about the sparsity of event occurrences. Specifically, we improve on a previous heuristic by implementing heuristic search algorithms and test reinforcement learning algorithms on the task. Finally, we implement an U-Net for predicting event occurrences that are caused by spatial relations between environment features for modeling more realistic scenarios, improving upon detections per second by 77% and decreasing average detection time by 56% when compared to a uniform sweep strategy.