A Text Mining Application Using Weighted Majority Voting Ensemble Method

In text mining, sentiment analysis is gaining popularity day by day although it has been recently introduced. One of the important feedback parameters of this research is the opinion about text-based content. The general goal in this aspect is to analyze product and service reviews or comments so that they can be compared and contrasted with each other via the ratings they get. An ensemble method which we have proposed earlier is used in this study to boost the classification accuracy of different conventional single machine learning models. Five analytical models that are related but not identical are implemented and their class decisions are integrated using a special weighted majority voting ensemble mechanism called WMVE to increase the classification score of the data mining technique. Naïve Bayes, OneR, Hoefding Tree, REPTree, and KNN methods are utilized as base classifiers in the ensemble and their class decision are integrated into the WMVE method. At the same time, outputs were compared to the ones obtained by Standard Majority Voting Ensemble (MV) including the same base classifiers. Based on the findings, the WMVE model demonstrated superior performance compared to other classifiers, achieving an average accuracy of 77.35 and F-Score of 77.19 values. Consequently, the ensemble model including WMVE is used to enhance sentiment analysis classification performance.