MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641101581 A) filed by Vignans Nirula Institute Of Technology And Science For Women on August 22, 2026, for Adaptive Hybrid Machine Learning System And Method For Real-Time Prediction Of Soil Moisture For Smart Irrigation.
Inventors include Srikanth Kilaru; K. V. S. S. Rama Krishna; Sk. Sharmila; V. Pavani; D. Amulya; T. Kameswari Smruthi; G. Triveni; and M. Dattasri.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The current invention introduces an Adaptive Hybrid Machine Learning System and Method for Real-Time Prediction of Soil Moisture for Smart Irrigation in order to address problems with traditional scheduled irrigation systems and prediction-based on a single algorithm used in machine learning known from prior art. The current invention captures environmental data, including such factors as temperature, humidity, precipitation, atmospheric pressure, soil pH value and concentration of soil nutrients, using IoT devices installed in fields and/or manually entered by users and then performs a pre-processing step, which includes missing value filling, normalization and standardization of collected data. The plurality of independent base machine learning engines, including Linear regression, Random forest regression, Support vector regression and Artificial neural network, predicts soil moisture based on processed data, while the Random Forest regression engine additionally evaluates feature importance ranking in order to improve input features representation for the Artificial neural network engine. A hybrid ensemble aggregation engine is used to combine the results obtained from the base engines to produce a final and accurate soil moisture prediction result, and from which the irrigation recommendation engine calculates the water requirement and produces the irrigation recommendation, which can be displayed via a web interface and/or communicated to the IoT-based actuation subsystem for automation of irrigation. The explainability module displays feature importance results to help build user trust, whereas the continuous/adaptive learning module periodically re-trains the algorithm based on new field data to maintain its accuracy. Experimental results show that the proposed hybrid algorithm offers better prediction accuracy, computational efficiency, robustness, scalability, and water conservation than current smart irrigation algorithms.
Disclaimer: Curated by HT Syndication.
