Author identification based on the ensemble learning approach supported by optimization-based feature selection methods

Accessing accurate information has been reduced by increasing opportunities for searching, copying and disseminating data via the Internet and especially social media. One of the studies conducted in the field of text mining in obtaining correct information from data is the author identification. A text carries the characteristics of the person who wrote it, and these properties can be used to identify the author of the text. In our study, a collection contained 54 authors, and a total of 46 837 texts with different numbers and different sizes, was used. A mixed analysis was prepared by combining two different analyzes and two analyzes to reveal the characteristics of the author. In order to increase the efficiency of the analysis results two different feature selection methods were offered, based on the combination of Random Forest Algorithm with Genetic Algorithm and Chicken Swarm Optimization Algorithm. As a result of the analysis, the most efficient author identification solution was supported with Ensemble Learning Algorithms. The best performance was achieved as 95.74% in the use of Decision Tree as a classifying method in Bagging Algorithm on a data set with ten authors, which was created as a result of operations performed with the Mixed Analysis and afterwards with the Genetic Optimization algorithm.