MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202611064397 A) filed by Noida Institute Of Engineering And Technology Niet on May 21, 2026, for Noise-Robust Financial Forecasting System With Gated Sentiment-Temporal Fusion Architecture.

Inventors include Dr. Varsha Jotwani; and Khushboo.

The application for the patent was published on July 17, 2026, under issue no. 29/2026.

Abstract: A financial forecasting system (100) for generating next-period price predictions integrates a sentiment signal generation processor (101), a dynamic gating fusion processor (102), and a temporal prediction engine (103). The sentiment signal generation processor (101) processes unstructured financial text through a Transformer encoder (104) and classification projection layer (105) to produce a three-element one-hot directional trend vector (106) and a scalar confidence coefficient (107). The dynamic gating fusion processor (102) concatenates structured numerical price features with the trend vector (106) and confidence coefficient (107), computes a gating scalar through a sigmoid activation unit (109), and applies element-wise multiplication and complementary accumulation to produce noise-attenuated fused feature vectors (112). The temporal prediction engine (103) applies multi-head self-attention to the fused vectors to generate price predictions (114). The gating mechanism adaptively regulates sentiment signal contribution based on confidence, providing noise resilience. The system is applicable to equity market forecasting on regulated exchanges.

Disclaimer: Curated by HT Syndication.