MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641089482 A) filed by Mrs. Sreedevi Kadiyala; Mr. Chandra Srinivas Potluri; Mothukuri Sai Prasad Rao; Ravindra Changala; K. Bhargavatriveninandana; K. Nagamani; Dr. Annapurna Gummadi; and Dr. Mahesh Kotha on July 22, 2026, for Optimizing Web Mining Through Enhanced Pattern Recognition And Intelligent Learning Models.
Inventors include Mrs. Sreedevi Kadiyala; Mr. Chandra Srinivas Potluri; Mothukuri Sai Prasad Rao; Ravindra Changala; K. Bhargavatriveninandana; K. Nagamani; Dr. Annapurna Gummadi; and Dr. Mahesh Kotha.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: ABSTRACT OF THE INVENTION: A system and method for optimizing web mining through enhanced pattern recognition and intelligent learning models. The present invention provides a multi-layered, self-adaptive architecture for efficient and accurate analysis of heterogeneous web data. The system comprises a data acquisition layer, a pre-processing module, and a core processing engine. The core processing engine features a novel adaptive feature extraction unit that employs 'Adaptive Principal Component Analysis (A-PCA)' to dynamically reduce data dimensionality while preserving critical information. This is complemented by a semantic context analyzer that uses a knowledge graph to understand the contextual relationships between data elements, thereby mitigating issues of data ambiguity. The extracted features are processed by a unique dual-stage intelligent learning model. The first stage is an ensemble of a deep neural network (DNN) and a support vector machine (SVM) that work in parallel to perform primary pattern recognition. The second stage is a reinforcement learning (RL) agent, which functions as a real- time optimizer. The RL agent continuously monitors the performance of the primary classification models and the evolving structure of incoming data. Based on this feedback, it dynamically adjusts the parameters of the A-PCA feature extractor and the weights of the ensemble classifiers to ensure sustained optimal performance. This creates a self-optimizing loop that allows the system to adapt to the dynamic nature of web content without human intervention. The system is designed for scalable, distributed computing, making it suitable for high-volume data streams. This invention significantly improves the efficiency and accuracy of web mining, finding applications in personalized recommendations, market intelligence, and fraud detection, thus overcoming the limitations of static, non-adaptive prior art systems.
Disclaimer: Curated by HT Syndication.
