publications

2022

1. Springer

   Approaches for Improving the Performance of Fake News Detection in Bangla: Imbalance Handling and Model Stacking

   In Proceedings of International Conference on Fourth Industrial Revolution and Beyond 2021. 2022

   Abs HTML

   Imbalanced datasets can lead to biasness in the detection of fake news. In this work, we present several strategies for resolving the imbalance issue of fake news detection in Bangla with a comparative assessment of proposed methodologies. Additionally, we propose a technique for improving performance even when the dataset is imbalanced. We applied our proposed approaches to BanFakeNews (Hossain et al. in Banfakenews: A dataset for detecting fake news in bangla. arXiv preprint arXiv:2004.08789, 2020), a dataset developed for the purpose of detecting fake news in Bangla comprising 50K instances but is significantly skewed, with 97% of the majority instances. We obtained a 93.1% F1-score using data manipulation techniques such as SMOTE and a 79.1% F1-score using without data manipulation approaches such as Stacked Generalization. Without implementing these techniques, the F1-score would have been 67.6% for baseline models. We see this work as an important step towards paving the way of fake news detection in Bangla. By implementing these strategies the obstacles of imbalanced dataset can be removed and improvement in the performance can be achieved.

2023

1. IEEE

   A Study on Future Lockdown Predictions using ANN

   In International Conference on Next-Generation Computing, IoT and Machine Learning (NCIM) 2023

   Abs HTML

   More than three years have passed since the first detection of COVID-19, and our generation had no clue as to what was coming or how dangerous the disease was. Unfortunately, this is not the end of such diseases, and in order to minimize losses, we may have to deal with many more lockdowns over new diseases. With the data from the recent pandemic, we investigated how to anticipate the potential lockdown date in the future. We took into account the first detection, the first death by COVID-19, the WHO emergency declaration, the total GDP of the various nations, the GDP growth, the population density, etc. This paperwork predicts the day passed after the first COVID-19 positive detection inside the country when the country’s government declared a lockdown (FDL) and the day when the WHO announced COVID-19 as an emergency of international concern (WEL). This research uses a variety of machine learning models, including a modified Artificial Neural Network (ANN), for the prediction. This study presents the successful tuning and training of an ANN model using the lockdown dataset, ultimately achieving an impressive accuracy rate of 96% for FDL (First Detection Lockdown) and 98% for WEL (World Emergency Lockdown), which is more accurate than any other applied traditional and/or tree-based machine-learning models.