Chicken Algorithm based Feature Selection for Arabic Information Searching System

Abstract

Because of the huge amounts of documents that exist on the internet, digital libraries, and e-mails, finding the related document becomes highly important and required nowadays. Generally, this process is implemented after implementing feature selecting approach which selects the proper features for enhancing the accuracy of searching. Feature selection can be defined as a method that is utilized in various application for eliminating redundant and irrelevant features. The dataset is simplified via this approach; via diminishing its dimensionality and recognizing related features with no decrease in the accuracy of prediction. Generally, datasets have huge dimensionality, as the learning algorithm could not function properly before eliminating these unrelated features. The running time regarding the learning algorithm is reduced significantly when the number of unrelated features is reduced. The majority of the selection of the feature depends on the frequency-inverse document frequency (TF-IDF) values, which are usually not effective, and English is the main focus of the research to retrieve the various files. In this research, the method of selection of feature, which is based on the chicken swarm algorithm, is used to solve many different problems, but it is not used in the selection of feature in Arabic information retrieval . In order to validate this technique, the accuracy of the evaluation including the comparison with two algorithms of swarm algorithms has been adopted. In addition, a group of experiments were conducted on an Arabic corpus(NLEL).The proposed system has effectively evaluated the algorithm and improved its accuracy in the system of retrieving the Arabic information.