A Deviant Load Shedding System for Data Stream Mining

Abstract

Load shedding is imperative for data stream processing systems in numerous functions as data streams are susceptible to sudden spikes in volume. The proposed system is an attempt to seek and resolve four major problems associated with data stream, which include load shedding and anti-shedding time, number of transactions pruned and selecting predicate; using efficient mining system. The frequent pattern discovered in data stream used in the model exploits the synergy between scheduling and load shedding. This paper also proposes various load shedding strategies which reduce and lighten the workload of the system ensuring an acceptable level of mining accuracy using various parameters like transaction, priority and attributes of data mining. A majority chunk of workload in mining algorithm lies in the innumerable item sets, which are counted and enumerated. The approach is based on the frequent pattern matching principle of stream mining which involves reducing the workload to maintain smaller item sets.