Opinion Mining Using Hybrid Lexicon-Based Approach and Machine Learning Techniques for Political Security Threat

One of the elements of the national security domain is political security that associated with risks and threats such as riots and civil war that could potentially jeopardize the social stability of a nation. The complexity of political security surged with the advancement of personal devices and internet connectivity contribute to massive online information sharing. The information may consist of sentiments that may pose risk to political security. Thus, maintaining political security currently require an enhanced mechanism explicitly designed to monitor sentiments or opinions, and detecting any extreme emotions triggered online that can lead to negative effect as well as threats. To fill that gap, this thesis presents a theoretical framework in predicting political security threats in cyberspace by combining a lexicon-based approach and machine learning. For machine learning threat classifier, three (3) machine learning techniques which are Naïve Bayes, Support Vector Machine, and Decision Tree were tested. The result was evaluated using performance assessment. The comparison of accuracy, precision, and recall was presented to validate the hybrid approach: lexicon-based approach and machine learning to be applied in the proposed framework. The evaluation result demonstrated the accuracy value of classification and prediction using the combination of the lexicon-based approach and Decision Tree was higher than other approaches and provided the optimum result for this framework. This framework offers valuable insights in opinion mining realm to predict threats focusing on political security elements and demonstrate the definite relation of emotion, sentiment with political security threats