AI-Powered Prediction System Using NLP and Deep Learning
International Journal of Advanced Research in Science Communication and Technology · 21.05.2026 · 149
DOI: 10.48175/ijarsct-35622 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Research in Science Communication and Technology, online erschienen am 21.05.2026. Verfasst von Dr. Pankaj Agarkar und 3 weiteren Autor:innen. Die Arbeit nennt 48 Literaturangaben.
Abstract (Originalsprache)
Prediction systems grounded in textual data have become indispensable across high-stakes domains including clinical decision support, financial signal detection, and digital misinformation analysis. Classical statistical approaches and shallow machine learning methods have demonstrated satisfactory performance on narrow, well-curated datasets, but they struggle to generalise once input distributions shift or domain vocabulary diverges from training corpora. Deep learning, and more specifically the pre-trained transformer paradigm, has substantially narrowed this gap; nevertheless, single-architecture solutions routinely leave accuracy on the table when applied to tasks that demand both rich contextual encoding and explicit sequential reasoning. This paper presents a cohesive, end-to-end AI- powered prediction framework that fuses BERT- derived contextual representations with a two-layer bidirectional LSTM (BiLSTM) classification head augmented by an additive attention mechanism. The system is designed as a modular pipeline: text acquisition and normalisation, augmentation-based imbalance handling, deep encoding, sequential modelling, and post-hoc probability calibration are treated as independent, replaceable stages. Experimental evaluation across three publicly available benchmark datasets — the LIAR fake news corpus, Stanford Sentiment Treebank v2, and a health-claim verification collection — confirms that the hybrid BERT-BiLSTM-Attention architecture outperforms five competitive baselines on macro-averaged F1 and area under the ROC curve. Ablation experiments quantify the individual contributions of the attention layer, recurrent head, augmentation strategy, and temperature scaling. A discussion of deployment trade-offs addresses inference latency, continual adaptation, and algorithmic fairness..
Zitieren
APA
Agarkar, D. P., Zalte, T., Shikalgar, F., & Kardile, U. (2026). AI-Powered Prediction System Using NLP and Deep Learning. International Journal of Advanced Research in Science Communication and Technology, 149. https://doi.org/10.48175/ijarsct-35622
Vancouver
Agarkar DP, Zalte T, Shikalgar F, Kardile U. AI-Powered Prediction System Using NLP and Deep Learning. International Journal of Advanced Research in Science Communication and Technology. 2026:149. doi:10.48175/ijarsct-35622
BibTeX
@article{Agarkar2026,
author = {Agarkar, Dr. Pankaj and Zalte, Tushar and Shikalgar, Farhan and Kardile, Umesh},
title = {AI-Powered Prediction System Using NLP and Deep Learning},
journal = {International Journal of Advanced Research in Science Communication and Technology},
year = {2026},
pages = {149},
doi = {10.48175/ijarsct-35622}
}
EUVIMED