AccuMediaNet: Ensemble Deep Learning Model based Media Accountability and Reporting

The increasing prevalence of biased or unethical media reporting, particularly in sub-judice criminal cases, poses a significant threat to judicial neutrality and public perception. This study introduces AccuMediaNet, an ensemble deep learning model that combines AlexNet and Transformers with attention mechanisms to evaluate media accountability effectively. The model leverages semantic and syntactic features, alongside metadata such as source credibility and journalist identity, to classify content into accountable and non-accountable categories. Comprehensive evaluations across training, validation, and test datasets yielded a remarkable accuracy of 97.55%, supported by precision (96.72%), recall (97.00<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), and F1-score (96.86<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). The model outperforms traditional approaches like SVM and standalone CNNs or Transformers, as evidenced by its superior AUC of 98.15%. Error analysis revealed that misclassification of neutral content and subtle sensationalism were the primary challenges, suggesting areas for refinement. The integration of explainable AI ensures transparent predictions, making the model a reliable tool for journalists, legal professionals, and media regulators. AccuMediaNet provides a scalable, efficient, and ethical solution to uphold accountability in media reporting, paving the way for more responsible journalism.