期刊名称：International Scientific Journal of Management Information Systems
印刷版ISSN：1452-774X
出版年度：2013
卷号：8
期号：1
页码：008-019
出版社：University of Novi Sad
摘要：This contribution presents an analysis on two key approaches to studying the characterization of English alphabets in doctoral thesis. The approaches (rescale range analysis (RSA) and fractal characterization (FC)) are discussed from the simulation point of view. For RSA, Hurst exponent value was used for the string of English alphabets in composing doctoral abstracts of engineering-based research work. FC involves the combined application of Cantor dust knowledge and fractal box dimension estimate by box counting and probability. For the two approaches, four engineering-based doctoral abstracts were studied with the total length of 512 alphabets in each case. The average computerized rescale range value was found to increase with the increase in data length for all cases. The Hurst exponent values for all cases distinctively range between 0.4146 and 0.4873 (i.e. negative correlation). The relative percentage error computed for the estimated fractal box dimension of Cantor Dust when compared with the literature result was 15.7% (i.e. the algorithm used in this study for the estimate will tolerate maximum of 15.7% error for any study case): Comparisons of sorted alphabets by frequency and estimated fractal box dimension for four abstract cases range between 37% to 77% agreement. The average percentage agreement among the four cases sorted by frequency was 31.5% and the average was 43% for sorting done by estimated fractal box dimension (due to recognition of placement and timing of usage of the English language alphabets in the studied cases). The graphs of estimated probability and fractal box dimension distribution for the studied cases follows trend.
关键词：fractals characterization; Hurst exponent; fractal box; Cantor Dust; rescale range
