Citation

Johra, Fatema Tuj and Nisa, Syed Qamrun and Khan, Mohammad Shadab (2026) A Comparative Analysis of Sentiment Detection Using Classical Machine Learning and Deep Learning Approaches. Advances in Science, Technology and Innovation. pp. 147-152. ISSN 2522-8714

Text A Comparative Analysis of Sentiment Detection Using Classical Machine Learning and Deep Learning Approaches _ Springer Nature Link.pdf - Published Version Restricted to Repository staff only Download (713kB)

Abstract

Sentiment analysis converts unstructured social text into decision-ready signals by classifying posts as positive, neutral, or negative. We present a deployment-ready pipeline for tri-class detection that unifies rigorous preprocessing, class-balance handling, feature construction, and a reproducible evaluation protocol. A single scaffold supports two methodological families—lightweight, interpretable techniques, and context-aware sequence encoders—so comparisons are fair and components are swappable without service disruption. To meet production constraints, the system emphasizes low-latency inference, calibrated decision thresholds, audit-friendly diagnostics, and safeguards for data and label drift, including periodic refresh routines. Dashboards expose precision and recall patterns and common error modes to guide remediation while maintaining consistent behavior across domains. The framework is domain-portable, minimizes overfitting to any one stream, and documents implementation choices for governance and reproducibility. In aggregate, the work distills the end-to-end steps required to deliver consistent, transparent, and scalable tri-class sentiment analysis for real-time or near-real-time applications, while clarifying trade-offs between interpretability and contextual nuance and providing an extensible path for future enhancements. Evaluation across varied short-form text sources confirms stable gains over simple baselines, with improved F1 under class imbalance and consistent generalization to held-out domains.

Item Type:	Article
Uncontrolled Keywords:	Machine learning
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	01 Oct 2026 01:15
Last Modified:	01 Oct 2026 01:15
URII:	http://shdl.mmu.edu.my/id/eprint/16742

Downloads

Edit (login required)