Abstract-Despite the recent emergence of research, creating an evolving fuzzy clustering method that intelligently copes with huge amount of data streams in the present high-speed networks involves a lot of difficulties. Several efforts have been devoted to enhance traditional clustering techniques into on-line evolving fuzzy able to learn and develop continuously. In line with these efforts, we propose kEFCM, kNN-based evolving fuzzy clustering method. kEFCM overcomes the problems of computational cost, dynamic fuzzy evolving, and clustering complexity of traditional kNN. It employs the least-squares method in determining the cluster center and influential area, as well as the Euclidean distance in identifying the membership degree. It enhances the traditional kNN algorithm by involving only cluster centers in making classification decisions and evolving on-line the clusters when a new data arrives. For evaluation purpose, the experimental results on a collection of benchmark datasets are compared against other well-known clustering methods. The evaluation results approve a good competitive level of kEFCM.

No abstract

In this paper we considered a theoretical evaluation of data and text compression algorithm based on the Burrows-Wheeler Transform (BWT) and General Bidirectional Associative Memory (GBAM). A new data and text lossless compression method, based on the combination of BWT1 and GBAM2 approaches, is presented. The algorithm was tested on many texts in different formats (ASCII and RTF). The compression ratio achieved is fairly good, on average 28-36%. Decompression is fast.