Identification of Extreme Guilt and Grave Fault in Bengali Language using Machine Learning

Abstract

Though huge amount of study has been done on the Bengali Language for information retrieval, but none of them deals with extreme guilt ( ) and grave fault ( ) in the Bengali Language. In this study, we have described extreme guilt ( ) and grave fault ( ). We have used three machine learning methods, such as Logistic Regression (LR), Support Vector Machine (SVM) and Multinomial Naive Bayes (MNB) as the baseline classifiers among the baseline classifier, MNB shows the accuracy of 89%. Ensemble learning has been used to improve the baseline classifiers. We have implemented an Ada Boost algorithm and Maximum voting classification decision method depending on the results of baseline classifiers. Maximum voting and Ada-Boost algorithms have shown an accuracy of 91% and 92% respectively. We have modified the Ada-boost algorithm using Principal Component Analysis (PCA) and named it JR-Ada-Boost. It outperforms all algorithms and gives an accuracy of 94%.