Optimization of multi-class document classi_cation with computational searchpolicy

Abstract

In the era of internet communication, many electronic documents are distributed via website every split second. The research interest for the process of knowledge discovery has changed from working with traditional printed documents to processing online data such as online news document classifica-tion. Most of the online data is in text documents and therefore the optimization of multi-class document classification is becoming a challenge for society today. Traditional search policy for the feature selection process is degrading with exhaustive search for complex features in document classifica-tion. Therefore, meta-heuristic based computational search is also becoming good solution to overcome the problem of exhaustive search. The search policy of a computational algorithm can provide the global optimal solution using a random search approach and selected optimal features can support finding the optimal classification results. In this paper, Cuckoo optimization (CO), Firefly optimization (FO), and Bat optimization (BO) algorithms are observed to overcome the problem of multi-class document classification by applying their adaptive search policies for searching for the optimal feature subset in a feature selection process. In addition, J48 and support vector machine (SVM) classifiers are used to evaluate the quality of selected feature subsets. The results from the proposed system are compared with traditional Best First search (BFS) and Ranker search (RS) based classification results. Furthermore, the results analysis is performed using various measure-ments from the point of view of performance analysis and complexity cost. According to the experimental results, the proposed system can generate the good multi-class document classification results using our computational search policy.