Advancing physical sciences through scholarship, publishing, and professional fellowship.

𝗔𝗻 𝗘𝘅𝗽𝗹𝗮𝗶𝗻𝗮𝗯𝗹𝗲 𝗖𝗡𝗡–𝗕𝗶𝗟𝗦𝗧𝗠–𝗥𝗮𝗻𝗱𝗼𝗺 𝗙𝗼𝗿𝗲𝘀𝘁 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝗳𝗼𝗿 𝗘𝗽𝗶𝗱𝗲𝗺𝗶𝗰 𝗪𝗮𝗿𝗻𝗶𝗻𝗴-𝗦𝗶𝗴𝗻𝗮𝗹 𝗮𝗻𝗱 𝗗𝗶𝘀𝗲𝗮𝘀𝗲-𝗦𝘁𝗮𝘁𝘂𝘀 𝗖𝗹𝗮𝘀𝘀𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻

A recent study published in the Journal of the Nigerian Society of Physical Sciences evaluated an explainable CNN–BiLSTM–Random Forest framework for two separate epidemic-related classification tasks: identifying symptom-bearing Ebola warning signals in tweets and classifying COVID-19 patient records as SARS-CoV-2 positive or negative. On the social media task, the revised framework achieved 0.918 sensitivity, 0.975 precision, and an F1-score of 0.946, with most of the improvement coming from expanding the tokenizer vocabulary and properly training the feature extractors. On the clinical task, a Random Forest using laboratory variables performed best, while adjusting the operating threshold raised sensitivity to 0.711 at 90% specificity and 0.803 at 80% specificity. The authors therefore position the clinical model as a post-haematology, pre-PCR triage aid rather than a screening or diagnostic tool.