Özet:

Abstract The number of traits that can be retrieved from the vast amounts of data of various forms available today is enormous. This is especially true for text data, which has benefited from the proliferation of multimedia applications. Using every available feature for each of the classification tasks can be not just time-consuming but also performance-detrimental. When a number of measurements, or features, have been acquired from a set of objects in a standard statistical pattern recognition problem, feature extraction is a frequent technique employed prior to classification. To achieve this, it is necessary to create a mapping from original representation space to a new space in which the classes may be more readily distinguished from one another. Selecting an appropriate feature set to characterize the patterns being classed is a common requirement in challenges involving knowledge discovery and pattern classification. This is because the classifier's performance and the cost of classification are both highly sensitive to the features used for the classifier's construction. To identify near-optimal solutions to such optimization issues, Genetic Algorithms (GA) present an appealing strategy. High-quality approaches to optimization and exploration issues can be quickly and easily generated using genetic algorithms, which rely on bioinspired operators including mutation, crossover, and selection. The genetic algorithm is an approach to solve optimization problems with constraints and without them, inspired by natural selection, the mechanism behind biological evolution. The genetic algorithm iteratively improves upon a pool of candidate solutions. A genetic learning and evolution model is used to pick or extract features while simultaneously designing a classifier. In this research Independent Subset Generation using Genetic Algorithm for Improved Classification (ISG-GA-IC) model is proposed for accurate selection of independent features for enhancing the classification levels. In this paper, we provide the results of our studies on the use of evolutionary algorithms for feature extraction and selection in high-dimensional data sets.

Anahtar Kelimeler: