ICM Compression System Depending On Feature Extraction

Abstract The goals of data compression is the task of providing space on the hard drive, and reduce the use of bandwidth in the transmission network and transfer files quickly. In this paper intelligent techniques are used as ways of lossless data compression. These methods applied using clustering techniques. Clustering is one of the most important data mining techniques. The proposed system presents a new algorithm used to determine the best compression method. This algorithm, called the ideal compression method (ICM). In addition to ICM system there are three clustering algorithms were used compress the databases and also we proposed new decompression algorithm was used to recover the original databases. The compression algorithms are different in the method of selecting attributes (which parameters are used as centers of clusters). These algorithms are improved K-means, k- mean with the medium probability and k-mean with maximum gain ratio. ICM used to determine the best one of these three algorithms that can be used to compress the database file. The main objective of this research is to select optimal compression method for each database. The ICM algorithm depends on the property of min-max removal and also depends on a number of conditions that used to determine the best compression method. This method continued in removing min-max column until remain one column. The residua column data was used to specify best compression method. The standard k-means algorithm suffering from several drawbacks such as it was dealt with only numerical data types, number of clusters needed to be specified by the used and the centers of clusters selected randomly. The three compression algorithms are modification algorithms to the standard kmeans algorithm. The modification was proposed in selecting the number of the clusters centers, specifying the number of the clusters and in dealing with the data types. Several experiments on different databases have been performed and the results are compared. The results shows that the maximum saving percentage is 98% minimum saving percentage is 61%, maximum decompression time is around 14 minutes, minimum decompression time is 6 seconds, maximum compression time is 17 minutes, minimum compression time is 5 seconds, maximum compression size is 1073 kilobytes and minimum compression size is 7 kilobytes. This research is organized as follow. Section one shows the introduction, Section two explains major clustering techniques, Section four shows the methodology of compression and decompression algorithms and system structure, Section five presents experiments and results and section six offers the conclusion.