Compression Schemes for Mining Large Datasets: A Machine Learning Perspective

Compression Schemes for Mining Large Datasets: A Machine Learning Perspective

M. Narasimha Murty, T. Ravindra Babu, S. V. Subrahmanya


This publication addresses the demanding situations of information abstraction iteration utilizing a least variety of database scans, compressing info via novel lossy and non-lossy schemes, and accomplishing clustering and type without delay within the compressed area. Schemes are provided that are proven to be effective either by way of house and time, whereas at the same time offering an identical or larger category accuracy. positive aspects: describes a non-lossy compression scheme in line with run-length encoding of styles with binary valued good points; proposes a lossy compression scheme that acknowledges a trend as a chain of positive aspects and making a choice on subsequences; examines even if the identity of prototypes and lines could be completed concurrently via lossy compression and effective clustering; discusses how one can utilize area wisdom in producing abstraction; reports optimum prototype choice utilizing genetic algorithms; indicates attainable methods of facing substantial information difficulties utilizing multiagent structures.

Show sample text content

Download sample