Makine Öğrenmesi Yöntemleriyle Gerçek Olmayan Tüketici Yorumlarının Tespiti

Abstract

The competition among companies has been considerably increased in the recent years due to the significant developments in online shopping of services and the widespread usage of e-commerce. The product reviews became a primary factor shaping the buyers' decisions. Due to this factor, product reviews created a marketing area for fake reviews about products and services. In this thesis, a model which uses a multiple classifier system has been proposed to identify the negative deceptive customer reviews and the validation has been performed on a dataset which consists of hotel reviews. The proposed model has a better performance than the best model reported in literature for this problem. In this model, five classifiers have been applied by using majority voting combination rule. These classifiers are libLinear, libSVM, Sequential Minimal Optimization (SMO), Random Forest and J48. LibSVM and libLinear are two different implementations of support vector machines.