Generating Fake News Detection Model Using A Two-Stage Evolutionary Approach

While fake news is morally reprehensible, irresponsible parties intentionally use it to achieve their goals by disseminating vulnerable and targeted groups. Machine learning techniques have been researched extensively detect news. On the other hand, evolutionary-based algorithms are now gaining popularity in research community. In this study, a two-stage evolutionary approach proposed generate optimize mathematical equation for detection. first stage, tree-based Genetic Programming (GP) algorithm used expressions correlations between language-independent (Lang-IND) features, extracted from Fake.my-COVID19 dataset, newly curated dataset mixed Malay - English language. The uniqueness of that formed basic arithmetic operators or include complex such as addition, multiplication, subtraction, division, square, abs, log1p, sign, square root, exponential together with Lang-IND features variables. Prior second stage approach, sensitivity analysis applied shorten best while maintaining F1-score performance. an Adaptive Differential Evolution (ADE), fine-tune model. experimental results conclude can be detection model learn predict using features. Results shows GP scores 83.23% on at tree depth 8. After fine-tuning performance increases 84.44%. outperforms baseline six commonly-used machine algorithms, Random Forest having highest 84.07%. also tested separately two unseen datasets different domain topic language achieves acceptable F1-scores.