MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641048529 A) filed by Mr. M Rajavel; Saai Jaswant; Y Saichandana; Dp Ramadevi; and Neethu Jimmy Joy on April 16, 2026, for Kafka Driven Real-Time Recommendation System For Streaming User Behavior.
Inventors include Mr. M Rajavel; Saai Jaswant; Y Saichandana; Dp Ramadevi; and Neethu Jimmy Joy.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The project presents a scalable and fault-tolerant architecture for real-time sentiment analysis of YouTube live chat, utilizing Apache Kafka for high-throughput data ingestion and Spark Structured Streaming for continuous, low-latency data processing. The system is designed to handle large-scale streaming data characterized by high velocity, multilingual content, and code-mixed language patterns, which are common in modem live chat environments. Chandray a an-3 serves as a real- world case study, enabling evaluation under realistic, high-engagement conditions during a globally significant live event. The architecture incorporates an end-to-end streaming pipeline, beginning with data ingestion, followed by preprocessing steps such as tokenization, noise removal, and normalization of multilingual and informal text. Feature extraction techniques, including TF-IDF and word r A embeddings, are applied to transform raw text into meaningful numerical representations. Multiple machine learning models are evaluated for sentiment classification, including Logistic Regression and Long Short-Term Memory (LSTM) networks. Experimental results indicate that Logistic Regression outperforms other models in this real-time setting, achieving approximately 99.2% accuracy with an average latency of around 260 ms per batch, making it highly suitable for time-sensitive applications. Although LSTM models capture contextual dependencies more effectively, they achieve a lower accuracy of about 85.9% and incur higher computational latency due to limited training data and increased model complexity. To ensure robustness and operational reliability, the system integrates Prometheus and Grafana for real-time monitoring and visualization of key performance metrics such as throughput, latency, CPU and memory utilization, and system health. This monitoring framework enables proactive detection of bottlenecks and ensures consistent performance under varying workloads. The project demonstrates the effectiveness of combining distributed streaming technologies with efficient machine learning models to deliver accurate, low-latency sentiment insights. It establishes a practical and scalable solution for real-time analytics in mission-critical scenarios, such as live events, digital marketing campaigns, and public opinion monitoring, thereby enhancing decisionmakingand user engagement through timely and .actionable insights. ,
Disclaimer: Curated by HT Syndication.
