Using XCS as a Prediction Engine in Data Compression
Mohsen Sharifi, Amir Aavani, Shahab Tasharrofi · 2007
XCS has been used in several fields as a prediction engine before; its population based nature enables XCS to generate sets of properly pruned classification rules.Further, unlike other population based algorithms, it learns using rewards.These properties encouraged us to use it as a predictor engine for lossless data compression.In the compression context, XCS can be used to find the hidden relations in files of the same type.So, we used it as a preprocessor before an entropy encoder, to remove the existing correlations between file's bits or symbols.Removing correlations causes entropy encoders to achieve higher rates of compression.The results support this conclusion.