Trully: News Article Credibility Checker Utilizing Machine Learning Classification Techniques

News serves a vital function in shaping public perception and guiding decision-making across various domains. However, the proliferation of fake news has increasingly contributed to misinformation and distorted viewpoints. This study addresses the challenge of fake news detection by developing a software solution with machine learning-based classification model. The development process was systematically guided by the CRISP-DM and Agile Scrum frameworks. A balanced, labeled corpus of $\mathbf{1 3, 5 6 4}$ Philippine news articles was prepared and preprocessed to support model training. Deep learning models, including MLP, CNN, and RNN-LSTM with GRU, were evaluated for their effectiveness in classifying news credibility. The RNN-LSTM with GRU architecture demonstrated superior performance with $\mathbf{9 2. 8 0 \%}$ accuracy score and was integrated into Trully Web Application, a platform designed for public use. A $\mathbf{1 0 0 \%}$ user acceptance testing (UAT) pass rate validated the application’s reliability and usability, offering an accessible tool for evaluating news credibility specifically in the Philippine context.