The discovery of frequent patterns from various datasets is a crucial aspect of modern data science. It helps to understand the relationship, dependence, structure, and diverse nature of data to take future actions like business decision making, catalog design, etc. Sometimes finding only frequent patterns is not enough as some items are less frequent but preserve high importance. For this reason, researchers have introduced the concept of weighted frequent pattern mining. Moreover, the situation becomes more critical when the dataset's size is enormous, and new data batches come at different velocities. Existing methods that have been proposed so far to handle such data streams are time-consuming and memory-hungry. This paper proposes an efficient sliding window-based approach that captures the data stream in a compact tree and mine weighted frequent patterns that overcomes both problems. We have conducted extensive experiments with the proposed method on different datasets. Experimental results exhibit significant improvement regarding both runtime and memory usage compared to existing approaches.