The processes of extracting knowledge structures for continuous, rapid records are known as the Data Stream Mining. The main issue in stream mining is handling streams of elements delivered rapidly which makes it infeasible to store everything in active storage. To overcome this problem of handling voluminous data the authors exposed a novel load shedding system using window based aggregate function of the data stream in which they accept those tuples in the stream that meet a criterion. Accepted tuples are conceded to another process as a stream, while further tuples are dropped. This proposed model conceivably segregates the data input stream into windows and probabilistically decides which tuple to drop based on the window function.