MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096470 A) filed by Dr Rema Gopalan; Cmr Institute Of Technology; Mathu K; Dr. Swati Mishra; Dr. Sandhya Kalale Srinivas; Dr Usha S; and Prof Prasanna Khadkikar on August 10, 2026, for A System And Method For Deep Neural Network On Demand Forecasting And Promotional Lift Prediction In Retail.
Inventors include Dr Rema Gopalan; Mathu K; Dr. Swati Mishra; Dr. Sandhya Kalale Srinivas; Dr Usha S; and Prof Prasanna Khadkikar.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: ABSTRACT A SYSTEM AND METHOD FOR DEEP NEURAL NETWORK ON DEMAND FORECASTING AND PROMOTIONAL LIFT PREDICTION IN RETAIL This invention discloses a system and method for integrated demand forecasting and promotional lift prediction in retail using deep neural networks. The system comprises four modules: data ingestion, feature engineering, deep neural network core, and decision support. The data ingestion module aggregates multi-source retail data including point-of-sale transactions, SKU attributes, promotional calendars, pricing, competitor data, weather, and macroeconomic indicators. The feature engineering module generates lag features, rolling averages, and promotional flags. The deep neural network core employs a hybrid architecture combining Temporal Fusion Transformer for time-series modeling and Multi-Layer Perceptron for promotional feature interaction, with attention mechanisms providing interpretability. The model simultaneously predicts baseline demand and promotional lift percentage. The decision support module converts predictions into actionable insights for inventory optimization and promotion planning. Validation on three years of retail data demonstrated 18.4% lower forecasting error and 26.7% higher promotional lift accuracy compared to baseline models. The invention enables retailers to reduce waste, improve service levels, and optimize marketing ROI.
Disclaimer: Curated by HT Syndication.
