Scalable Real-Time Sentiment Analysis on Massive Social Media Streams Using Parallel and Distributed Computing
International Journal of Electrical, Electronics and Computers · 2025
DOI: 10.22161/eec.106.1 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Electrical, Electronics and Computers, 2025. Verfasst von Hafsa Maryam und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The rapid growth of social media streams has intensified the need for scalable, low-latency sentiment analysis pipelines that can operate under high-volume, real-time constraints. This paper proposes a distributed framework built on Apache Spark for massive parallel processing of text streams and seamless integration of a fine-tuned large language model (LLM), Grok-4, for sentiment classification. The system employs micro-batch streaming, distributed tokenization, and GPU-accelerated model serving to achieve real-time inference at scale. Experiments conducted on a 10-node cluster using a synthetic dataset of 10,000 tweets, extended to 1.2 million streaming events, demonstrate substantial performance gains. Our approach achieves a 5.4× improvement in distributed training throughput and a 4.7× reduction in inference time compared with single-node baselines. The streaming pipeline sustains 2,100 tweets per second with an end-to-end median latency of 120 ms, satisfying real-time constraints for high-volume applications. The fine-tuned Grok-4 model attains 92.8% sentiment classification accuracy, outperforming conventional machine learning baselines by 8.5% absolute, while preserving high throughput. Comparative analysis shows the framework scales nearly linearly with increasing cluster size and maintains robustness against executor failures and network-induced delays. The results highlight the effectiveness of combining parallel and distributed computing with advanced LLM-based natural language understanding for high-frequency social data analytics. The proposed architecture provides a practical foundation for scalable deployments in domains such as public health surveillance, financial market monitoring, and real-time situational awareness systems.
Zitieren
APA
Maryam, H., & Farid, A. (2025). Scalable Real-Time Sentiment Analysis on Massive Social Media Streams Using Parallel and Distributed Computing. International Journal of Electrical, Electronics and Computers. https://doi.org/10.22161/eec.106.1
Vancouver
Maryam H, Farid A. Scalable Real-Time Sentiment Analysis on Massive Social Media Streams Using Parallel and Distributed Computing. International Journal of Electrical, Electronics and Computers. 2025. doi:10.22161/eec.106.1
BibTeX
@article{Maryam2025,
author = {Maryam, Hafsa and Farid, Ahmad},
title = {Scalable Real-Time Sentiment Analysis on Massive Social Media Streams Using Parallel and Distributed Computing},
journal = {International Journal of Electrical, Electronics and Computers},
year = {2025},
doi = {10.22161/eec.106.1}
}
EUVIMED